<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>Trust-Buster.com — Blog</title>
    <link>https://trust-buster.com/blog</link>
    <atom:link href="https://trust-buster.com/blog/rss.xml" rel="self" type="application/rss+xml" />
    <description>OT security, equipment access control, NIS2 / CRA compliance.</description>
    <language>en</language>
    <lastBuildDate>Wed, 11 Nov 2026 18:00:00 GMT</lastBuildDate>
    <item>
      <title>A Wi-Fi Honeypot That Grants Nothing</title>
      <link>https://trust-buster.com/blog/a-wi-fi-honeypot-that-grants-nothing</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/a-wi-fi-honeypot-that-grants-nothing</guid>
      <pubDate>Wed, 11 Nov 2026 18:00:00 GMT</pubDate>
      
      <description>Against a Flipper Zero that probes, impersonates or forces a disconnect, here is the honeypot that grants nothing, what it sees and what it refuses to do.</description>
      <content:encoded><![CDATA[<p>A wireless network that exists only to be attacked has nothing left to protect, and that is what makes it a good sensor, so we set up a honeypot that grants nothing, in the form of a fake access point which advertises its name but accepts no connection, and its only job is to make visible the gestures of someone who would take too close an interest in the Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>).</p>
<p><em>This article follows the one we wrote about <a href="/blog/jamming-relays-and-the-flipper-zero">the Flipper Zero and radio detection</a>, but it goes down a level toward the concrete, on a case we really tested on the bench before writing about it.</em></p>
<h2 id="four-gestures-and-one-goal">Four gestures and one goal</h2>
<p>A consumer wireless attack tool has only a small repertoire of gestures against a network, and they look much alike from a distance, so it is worth naming them one by one because each leaves a different trace.</p>
<p>First there is reconnaissance, because the attacker listens and asks before anything else, and it calls out to find whether a given network is present, which your phone does too, but a probe that asks by name for an administration network where no such thing should exist is already information.</p>
<p>Then there is the fake network, or evil twin, because rather than entering yours the attacker can advertise their own under the same name and hope a device or a person connects to it by mistake, and its signature is simple to state because it is our network name broadcast from a transmitter that is not ours.</p>
<p>There is also the connection attempt, because the attacker can try to get in for real by starting the authentication and association exchange, and against our decoy this attempt never succeeds for a reason we give below, but the attempt itself, aimed at our access point, is what we want to see.</p>
<p>And finally there is the forced disconnect, because the attacker may seek not to enter but to evict, and deauthentication frames sent in bursts tear down existing links to force a reconnection one hopes to capture or simply to disrupt, and this attack looks like nothing at the radio level since the frames are valid, while it is entirely abnormal at the logical level.</p>
<h2 id="a-honeypot-that-grants-nothing-rather-than-a-fortress">A honeypot that grants nothing rather than a fortress</h2>
<p>Faced with these gestures the instinct is to harden a real door, but we preferred to set up a door that is not one. The Cube hosts a decoy network whose name is chosen for each site, for instance a technical-looking name. The access point must be declared to the company&#39;s Wi-Fi intrusion detection system. The decoy is protected by a key no one knows, because it is drawn at random at start-up, it is long, and it is never shown nor transmitted nor written in the clear anywhere but in the volatile memory of the chip that produced it, so no legitimate device has to connect to it and no attacker will get in for want of the key, and this network has nothing behind it, no service and no configuration and no access to anything. Its usefulness is not to be a fortress but a wire stretched across the corridor, and whoever approaches the Cube by the wireless path trips over it and we know.</p>
<h2 id="what-it-sees-and-what-it-does-not-do">What it sees and what it does not do</h2>
<p>The honeypot watches its channel, as a wireless intrusion detection sensor, or WIDS, would, and recognises the four gestures described above, so a probe that asks for the decoy&#39;s name, an authentication or association attempt aimed at it, the advertisement of its own name by another transmitter, and a burst of disconnects beyond a threshold, and each of these observations produces a timestamped event that travels to the Cube&#39;s trust anchor, the part that makes the decisions, over the same signed internal link as the rest of the system.</p>
<p>We also have to say what this mechanism does not do, because that matters just as much. It never grants access, because the decoy is a sensor and not an entry point, and it stays one even when it detects an attempt. It emits no forged frame, because it broadcasts only the ordinary signals of an access point that exists, and it fabricates no fake deauthentication and no fake probe and nothing that would resemble an attack turned back on the attacker. And it does not retaliate, because jamming or deauthenticating in return would be outside our role and unlawful on these bands in most countries, and we do not want to turn a sensor into a transmitter we would have to answer for.</p>
<h2 id="the-charge-port-as-a-sensor-and-not-a-door">The charge port as a sensor and not a door</h2>
<p>The same principle extends elsewhere in a less visible way, because the Cube charges through a USB-C connector and the two data wires of that connector, which no one uses to exchange anything, are read in read-only through a resistor and a clamp by the <span>Cube&#39;s guardian</span>, and this passive reading makes it possible to tell apart, without ever opening any dialogue, the charging cable we supply and a commodity adapter and a computer and an enumeration attempt such as a hostile peripheral plugged in place of the charger would produce. Here again the port grants access to nothing and becomes a sensor of its surroundings, and knowing that the Cube is no longer on its usual cable is information we would rather have.</p>
<h2 id="how-we-tested-it">How we tested it</h2>
<p>We are careful to describe only what we have seen work. Detection was first checked on the bench in a reproducible way, because the firmware carries a self-test that builds real wireless frames in memory and runs them through the same analysis path as frames received by radio, without emitting anything on the airwaves, and each of the four gestures triggers the expected alert there while a probe that asks for another network is correctly ignored, which shows that there is no obvious false positive. Then comes the real radio trial, where an outside bench, which is exactly the kind of tool one seeks to detect, produces the gestures while the Cube only recognises them, and the division of roles is clear because the tool attacks and the Cube observes, and nothing of that attacking tool appears in what we design or publish. The criterion we set ourselves is simple, because every gesture has to be logged and reported in under two seconds without any access having been granted.</p>
<h2 id="what-detection-proves-and-what-it-does-not">What detection proves and what it does not</h2>
<p>A honeypot that grants nothing and still lights up proves a gesture took place near the Cube but it does not prove an intent, because a probe may come from a distracted phone and a connection attempt from a misconfigured device, and that is why these events corroborate a decision without ever replacing it, since they feed a judgement made by the trust anchor. The limits deserve to be said plainly, because the decoy watches one channel at a time and spotting a fake network advertised on another channel calls for broader monitoring that still lies ahead of us, and like any radio device it is exposed to denial of service, since someone who wants to prevent rather than enter will manage by saturating the band, and the answer to that is organisational as much as technical.</p>
<h2 id="what-we-make-of-it">What we make of it</h2>
<p>We did not try to make the Cube impregnable by the wireless path, because that promise is never kept by those who make it, but we tried to make it impossible to approach quietly and to make the slightest attempt leave an exploitable trace reported where the decisions are made. A honeypot that grants nothing and a charge port that grants nothing are reliable sensors rather than attack surfaces precisely because of that absence of access, so we ship no offensive capability and no retaliation, and what we ship is a Cube that knows when it is being watched.</p>
]]></content:encoded>
    </item>
    <item>
      <title>The Site Manager&apos;s Rules: Alert, Suspend, Revoke</title>
      <link>https://trust-buster.com/blog/the-site-manager-rules</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/the-site-manager-rules</guid>
      <pubDate>Wed, 11 Nov 2026 11:25:00 GMT</pubDate>
      
      <description>The site manager&apos;s rules: who intervenes, when and from where, a window for the refrigeration contractor, graduated alerts and a log that keeps the proof.</description>
      <content:encoded><![CDATA[<p><strong>The site manager&#39;s rules, in brief</strong></p>
<ul>
<li>The Cube applies the site manager&#39;s rules. They state who may intervene remotely, when and from where. With a serial sentinel diode, they also bound the range of a setpoint.</li>
<li>Outside these rules, it alerts on three levels. It can suspend remote access, never the refrigeration, nor knowingly the alarms.</li>
<li>Every decision goes into a local, chained log.</li>
</ul>
<p><em>Sixth and last article in the series. The previous one, <a href="/blog/listen-without-being-able-to-speak">&quot;Listen without being able to speak&quot;</a>, described our sentinel diodes, which read the bus without being able to write to it. They see without cutting, but it remains to say who decides.</em></p>
<h2 id="the-site-manager-s-rules-one-sentence-each">The site manager&#39;s rules, one sentence each</h2>
<p>The manager of a store, a workshop, a plant or a warehouse generally has neither a security team nor time for a console, so each rule must fit in one sentence spoken aloud. For example: &quot;the refrigeration company may connect on weekdays from 7 am to 7 pm, from its usual address. It must not raise the defrost end temperature above 12 °C, nor the maximum defrost duration beyond 45 minutes.&quot; These thresholds of 12 °C and 45 minutes are only examples, to be validated by the refrigeration contractor. Time and duration can be checked without reading anything. The range, on the other hand, requires reading the setpoint, which is the job of the serial sentinel diode.</p>
<p>The Cube first proposes rules based on the behaviour observed over the first days. An anomaly already under way would then be learnt as normal, hence the prior check of the configuration described in <a href="/blog/observe-without-reading">&quot;Observe without reading&quot;</a>. The site manager then writes the rules with their refrigeration contractor, who knows the monthly or seasonal interventions, and a simple interface lets them adjust the rules alone afterwards.</p>
<p><strong>Revoking must be as simple as granting.</strong> Removing a technician means striking out a line. In the Cube, rules come down signed from the server, so a badly signed package is refused. Each package also carries a validity date, because a Cube cut off from the network could miss a revocation. Once that date has passed, it refuses all access and logs the fact as critical, and only break-glass then remains possible, until fresh rules return. The delay of a revocation is therefore bounded, at the cost of a general refusal if the Cube stays too long without news.</p>
<h2 id="the-intervention-window-at-night-too">The intervention window, at night too</h2>
<p>An on-call refrigeration contractor must be able to intervene at 3 am. Today, the secret often lies in a shared password. Since the attack, <a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">Dixell France</a> has asked for it to be replaced by named accounts with an expiry date. We go further: <strong>at night, the supervisor&#39;s password is no longer enough.</strong> The technician is enrolled through an app they agree to install, which is negotiated with each refrigeration company. The window then opens by agreement, and their supervisor credentials, still required, are useless outside the window.</p>
<p>The agreement can come from a person on site, whose badge or phone is recognised by the Cube, just as for authorising a machine. It can also come from the site manager or the on-call person: the phone rings, they see who is asking and for how long, then accept or refuse with a single gesture. The window closes by itself.</p>
<p><strong>With no network and nobody reachable</strong>, paper remains. The Cube then checks, offline, a break-glass QR code kept in a sealed envelope. Its use triggers a critical alert and leaves a signed receipt, handed to the server the next time an authorised phone passes by.</p>
<p><strong>When a technician leaves</strong>, their phone is removed and they can no longer open a window. Named accounts are still better than a shared account. A phone that is lent or stolen, and unlocked, does however open what its holder could open, because the Cube proves that an enrolled device was present, not who was holding it.</p>
<h2 id="three-alert-levels">Three alert levels</h2>
<p>An alert that rings for everything ends up no longer being read, hence three levels, which the site manager adjusts.</p>
<table>
<thead>
<tr>
<th>Level</th>
<th>Example</th>
<th>Proposed response</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Information</strong></td>
<td>Announced intervention, within a granted window.</td>
<td>Log and daily summary.</td>
</tr>
<tr>
<td><strong>Warning</strong></td>
<td>Request outside the time slot, or from an unusual address.</td>
<td>Immediate notification to the site manager, who accepts or refuses.</td>
</tr>
<tr>
<td><strong>Critical</strong></td>
<td>Setpoint beyond the permitted range, break-glass, several anomalies at once.</td>
<td>Immediate notification, and suspension of remote access if the rule provides for it.</td>
</tr>
</tbody></table>
<p>The range can only be seen when a sentinel diode reads the bus. An adjustment made in the supervisor itself, or on a LON bus, therefore escapes the Cube. The Cube ranks on these three levels its own events as well as the remote maintenance sessions reported to it by the satellites of articles 4 and 5.</p>
<h2 id="suspend-remote-access-never-the-refrigeration">Suspend remote access, never the refrigeration</h2>
<p>The suspension provided for by the site manager&#39;s rules targets a single thing, which is <strong>the door through which one enters the supervisor from outside.</strong> The controllers keep regulating and the alarms keep going out. At the <a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">Établissement français du sang</a>, the French blood service, an on-call technician and the alarms reportedly spotted the anomalies in time.</p>
<p>With the <a href="/blog/a-satellite-for-the-store-wi-fi">Wi-Fi gateway</a>, it is enough to stop letting the refrigeration contractor&#39;s tunnel through, or to close it if it is open. With the contractor&#39;s 4G router, the only safe hold is sometimes its power supply, which the Cube can control through a relay. But it is only cut with the contractor&#39;s written agreement, and only if the site manager has put this router on <a href="/blog/power-off-is-not-an-emergency-stop">the list of equipment that may be cut live</a>. <strong>If this router also carries the alarms, cutting it would silence them.</strong> They must first be given a separate path, for example an independent alarm transmitter. CERT-FR also advises <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">checking the insurance conditions</a> before cutting the data flow to the refrigeration contractor. Temperature recording remains <a href="https://info.agriculture.gouv.fr/boagri/instruction-N2006-8247/telechargement">mandatory in France</a> in frozen food storage premises, with records kept for at least one year. Cutting the data flow does not, however, remove local recording.</p>
<p>The Cube controls sockets and relays, quarantines a socket that someone tries to bypass, and logs every cut. The bounded opening of a supervisor&#39;s router reuses these same actions.</p>
<h2 id="what-the-log-proves-and-what-it-does-not">What the log proves, and what it does not</h2>
<p>Each entry is linked to the previous one through a cryptographic fingerprint, and the Cube signs the whole. Removing or changing an entry therefore breaks the chain visibly, and a loss is itself recorded, as detailed in <a href="/blog/a-proof-log-what-it-proves-and-what-it-does-not">&quot;A proof log&quot;</a>. The signing key is protected in a secure element.</p>
<p><strong>It proves</strong> that a window was opened, when, under which rule, with the agreement of which device and for how long, and which rule triggered a suspension.</p>
<p><strong>It does not prove</strong> what the technician did in the supervisor, unless a sentinel diode read the setpoints that were sent. Nor does anything in it say who was holding the phone. And its time is that of the Cube&#39;s clock. The log itself has limits, which <a href="/blog/a-proof-log-what-it-proves-and-what-it-does-not">&quot;A proof log&quot;</a> details, from sequence numbers to the chain head handed to the server. This log names technicians, and the badge or phone may also reveal the presence of employees, so it falls under the GDPR. The people concerned must be informed of it, and staff representatives consulted where required.</p>
<p><strong>Who answers for a wrongful cut?</strong> The site manager chooses the rule. The rule templates we propose and their application can be checked in the log, and the sharing of responsibilities is set out in the contract. An unjustified suspension does not stop the refrigeration, but during a genuine breakdown it delays the intervention, which may cost goods. It is better to write into the refrigeration contractor&#39;s contract how they obtain an emergency opening.</p>
<h2 id="three-questions-to-ask-your-refrigeration-contractor-tomorro">Three questions to ask your refrigeration contractor tomorrow</h2>
<p>These rules are only worth what is known about the installation, and three questions are enough to begin.</p>
<ol>
<li>How is our supervisor connected, through the site network, a 4G router, a built-in modem?</li>
<li>By which path do the alarms go out, and is it the remote maintenance path?</li>
<li>Who holds remote access today, and with which accounts?</li>
</ol>
<p>Your refrigeration contractor generally knows the answers. What remains is to know what to do when refusal itself becomes a risk, which we addressed in <a href="/iam-electric">&quot;IAM at the end of the power cord&quot;</a>.</p>
<p>The series: <a href="/blog/hacked-fridges-defrosting-from-afar">defrosting from afar</a>, <a href="/blog/why-the-fridges-stayed-open">why the fridges stayed open</a>, <a href="/blog/observe-without-reading">observe without reading</a>, <a href="/blog/a-satellite-for-the-store-wi-fi">a satellite for the store Wi-Fi</a>, <a href="/blog/listen-without-being-able-to-speak">listen without being able to speak</a> and <a href="/blog/the-site-manager-rules">the site manager&#39;s rules</a>.</p>
<p><em>Are you a site manager, a refrigeration contractor or a maintenance provider? We are looking for pilot sites, to be set up with your refrigeration contractor. Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>.</em></p>
<h2 id="sources">Sources</h2>
<ul>
<li><a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">CTOL, Dixell France guidance and end of support for XWEB EVO</a></li>
<li><a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">The EFS, quoted by Le Quotidien du Médecin</a></li>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a></li>
<li><a href="https://info.agriculture.gouv.fr/boagri/instruction-N2006-8247/telechargement">French Ministry of Agriculture, temperature recording for frozen food</a></li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration contractor, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Listen Without Being Able to Speak: Sentinel Diodes</title>
      <link>https://trust-buster.com/blog/listen-without-being-able-to-speak</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/listen-without-being-able-to-speak</guid>
      <pubDate>Wed, 11 Nov 2026 11:20:00 GMT</pubDate>
      
      <description>Listen without being able to speak: our sentinel diodes read the Ethernet and Modbus RTU of a refrigeration supervisor without being able to send it a command.</description>
      <content:encoded><![CDATA[<p><strong>Listen without being able to speak, in brief</strong></p>
<ul>
<li>Our Ethernet and serial sentinel diodes are satellites of the Cube. They listen to an existing cable, with no component able to transmit towards the refrigeration equipment.</li>
<li>On Ethernet, they see who talks to the supervisor, when and how much. On a Modbus RTU bus, they also read the setpoints sent to the controllers, so they can judge the value of a command.</li>
<li>They cannot disturb the installation with a command. But they cut nothing and may stay silent when they should not. Warranty and liability are therefore discussed before installation.</li>
</ul>
<p><em>Fifth article in the series. The previous one, <a href="/blog/a-satellite-for-the-store-wi-fi">&quot;A satellite for the store Wi-Fi&quot;</a>, described a gateway that places the supervisor behind a single controlled door, opened for a window decided by the site manager. One question remained: what can be seen when someone holding valid credentials, stolen or not, walks through that door and sends a command?</em></p>
<h2 id="listen-without-being-able-to-speak-the-principle">Listen without being able to speak, the principle</h2>
<p>Industry knows data diodes, which let information flow in one direction only, by hardware design, but their cost is not that of a store. We borrow the idea: to listen without being able to speak means reading what travels along a cable, while no component of the assembly can send anything back onto it.</p>
<p>The word diode is an image. These are not certified data diodes, which offer another level of assurance. Downstream, the Cube exchanges with an administration server, the operator&#39;s or ours depending on the contract, and that link carries only signed rules and logs, never a command towards the refrigeration. A sentinel diode sits on the existing cable, and what it observes travels up to the Cube by a separate path, in line with our habit of <a href="/blog/ot-retrofit-addons-instead-of-upgrades">adding rather than modifying</a>.</p>
<h2 id="on-the-supervisor-s-ethernet">On the supervisor&#39;s Ethernet</h2>
<p>The first diode goes on the cable linking the supervisor to its router, or to the gateway from the previous article. It relies on a <strong>passive Ethernet tap</strong>, which copies the signals to a monitoring port and <a href="https://www.garlandtechnology.com/blog/the-101-series-passive-network-taps">sends nothing back onto the monitored link</a>.</p>
<p>A fully passive copper tap, however, <a href="https://en.wikipedia.org/wiki/Network_tap">only works at 10BASE-T and 100BASE-TX</a>. On a gigabit link it therefore brings the negotiation down to 100 Mbit/s, which is enough for a supervisor. It delivers two streams, one per direction. Fitting it briefly interrupts the link, and when the supervisor is powered over the cable (PoE), this must be checked.</p>
<p>Over HTTPS, the diode sees who connects, when, for how long and with what volume, but not what was changed. That already gives the time and the duration, two signals described in <a href="/blog/observe-without-reading">&quot;Observe without reading&quot;</a>. On a readable protocol, plain HTTP or Modbus TCP, it also recognises setpoint writes.</p>
<h2 id="on-the-modbus-rtu-serial-line">On the Modbus RTU serial line</h2>
<p>Between the supervisor and the controllers there is often a serial line. The <a href="https://cars.uchicago.edu/software/epics/Modbus_over_serial_line_V1_02.pdf">Modbus over serial line specification</a> gives two-wire RS-485 as the most common, while RS-232 serves short point-to-point links. At Dixell, a brand of the Copeland group, <a href="https://www.hussmann.com/ns/Technical-Documents/XWi70K-EN.pdf">an XWi70K controller joins the Modbus RTU line through an external XJ485-CX module plugged into its HOTKEY port</a>. Some Danfoss installations use another bus, <a href="https://assets.danfoss.com/documents/211618/AJ430138910308en-000101.pdf">LON</a>, which we cannot listen to. Your refrigeration contractor will tell you which one equips your display cases.</p>
<p>The second diode connects to this RS-485 bus, although some RS-232 equipment requires a different receiver. Our assembly starts with an <a href="https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/3008/MAX3095,%2096.pdf">RS-485 receiver</a> with no transmit stage. Next comes an optocoupler, through which the signal passes as light, and then a microcontroller. An attacker in control of the microcontroller could make it report anything at all, but could not send a command to the controllers.</p>
<h3 id="from-the-bus-to-the-cube">From the bus to the Cube</h3>
<p>What the diode hears then reaches the Cube through a DIN rail module, the <a href="https://www.pusr.com/products/Serial-to-Dual-Band-WiFi-Converter.html">PUSR USR-DR164</a>. This RS-485 to dual-band Wi-Fi converter (2.4 and 5 GHz) also acts as a Modbus RTU/TCP gateway. It runs on 5 to 36 V and withstands -40 to 85 °C. On its own, however, it is bidirectional. The diode, reduced to its receive stage, therefore stays between the bus and the module, so that nothing can be transmitted towards the controllers, even by a compromised module.</p>
<p>It is wired in a daisy chain, input and output on its terminal block, like one more controller. It has neither termination nor biasing, since the specification reserves them for the ends of the bus and for a single point on it. The shield is carried through without being connected to earth. Its receiver presents only a quarter unit load, plus the leakage of its protections, so it weighs little on the bus. Series resistors protect its input. A bus that is terminated but not biased may oscillate when idle, which is measured at installation.</p>
<h2 id="reading-the-value-of-a-setpoint">Reading the value of a setpoint</h2>
<p>On this bus, the diode mainly reads writes: which controller receives which value, in which parameter, at what time. It tells reads from writes by their function codes, 05 and 06 (a single output or register), 0F and 10 (several), according to the <a href="https://www.modbus.org/file/secure/modbusprotocolspecification.pdf">Modbus protocol specification</a>. This is the <strong>value</strong> signal. It does, however, require the register map of each model, which not every manufacturer publishes.</p>
<p>The <a href="https://www.hussmann.com/ns/Technical-Documents/XWi70K-EN.pdf">XWi70K manual</a> gives three ranges. The defrost end temperature runs from -55 to 50 °C, the maximum defrost duration from 0 to 255 minutes, and the interval between two defrosts from 0 to 120 hours. The device accepts each value taken on its own, but it is their combination that makes a display case warm up. That combination is therefore what a rule watches, in the spirit of the FBI and CISA, who call for <a href="https://www.ic3.gov/CSA/2026/261105.pdf">defining reasonable setpoint ranges, with an alert when they are exceeded</a>. A combination never seen before on this display case, arriving at three in the morning, for instance, stands out at once.</p>
<blockquote>
<p><strong>Update of 11 November 2026.</strong> <a href="https://ransom-isac.org/blog/the-copeland-incident/">Ransom-ISAC</a> published on 6 November its analysis of a commercial refrigeration site attacked on the night of 2 November. The controller was reachable from the Internet with its factory administrator credentials, and the report counts 743 parameter writes across 31 devices, setpoints raised to 110 °C and 150 °C and several commands silencing the alarms, with no attribution. These bursts of writes are the kind of signal this article describes, and the report also recommends checking temperatures with independent instruments.</p>
</blockquote>
<h3 id="why-listen-to-the-bus">Why listen to the bus</h3>
<p>Nothing published so far names Modbus as the way in for the attacks. <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR</a> considers exposing management interfaces on the Internet contrary to good practice and asks for them to be withdrawn. We listen to this bus nonetheless, because it is where a setpoint reaches the controller, whatever path it took into the supervisor. On a LON bus, only the Ethernet diode then remains.</p>
<h2 id="what-this-proves-and-what-it-does-not">What this proves, and what it does not</h2>
<p><strong>What this proves.</strong> A sentinel diode designed without any transmitting component can neither write nor, therefore, disturb the installation with a command, even when compromised. It also provides a witness outside the supervisor, which matters when the supervisor itself is the one compromised. At the EFS, the French blood service, <a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">the incident was reportedly spotted by an on-call technician, and the alarm systems helped preserve the stocks</a>.</p>
<p><strong>What this does not prove.</strong> The diode cuts nothing. It reports to the Cube, which alerts the site manager. It is the gateway or a controlled socket that suspends incoming remote access. If the router being cut also carries the alarm reporting, the alarms first need a separate path. Suspension never targets the refrigeration.</p>
<p>A <strong>false negative</strong> remains possible. The diode only hears its own cable. An attack under way during the learning period would be learnt as normal. And a compromised supervisor could forge false controller responses, which is why an independent temperature probe remains useful on a critical site. An adjustment by the refrigeration contractor, announced within a granted window, triggers no alert and goes into the log and the daily summary. Outside a window, or beyond the permitted range, the alert is immediate.</p>
<h2 id="warranty-contract-and-liability">Warranty, contract and liability</h2>
<p><strong>The manufacturer&#39;s warranty.</strong> Can a third-party device connected to the bus be invoked to refuse a claim? The answer depends on each manufacturer.</p>
<p><strong>The refrigeration contractor.</strong> They must know that a device is listening and accept it. Ideally they would fit the diode themselves and offer it to their customers. Nothing changes in their interventions, and the witness also shows that their adjustments were within the rules.</p>
<p><strong>The insurer.</strong> The question deserves to be put to them. An independent witness can help document a loss, without guaranteeing that it will be covered.</p>
<p><strong>Whoever installs it.</strong> If the device disturbs the bus through an electrical fault, who is answerable? And if it saw nothing, who is at fault? Connection inside a cabinet remains the business of a refrigeration professional.</p>
<p>On these points, we recommend a legal review before any deployment.</p>
<p><strong>The last article in the series, <a href="/blog/the-site-manager-rules">&quot;The site manager&#39;s rules&quot;</a>, describes what the Cube does with these signals:</strong> intervention windows, graduated alerts, suspension of remote access and the log.</p>
<p><em>Do you maintain refrigeration supervisors or Modbus buses, and see other constraints? Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>.</em></p>
<h2 id="sources">Sources</h2>
<ul>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a></li>
<li><a href="https://www.ic3.gov/CSA/2026/261105.pdf">FBI / CISA, advisory of 5 November 2026 (PDF)</a></li>
<li><a href="https://ransom-isac.org/blog/the-copeland-incident/">Ransom-ISAC, analysis of the Copeland incident (6 November 2026)</a></li>
<li><a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">The EFS, quoted by Le Quotidien du Médecin</a></li>
<li><a href="https://www.garlandtechnology.com/blog/the-101-series-passive-network-taps">Garland Technology, how passive network taps work</a></li>
<li><a href="https://en.wikipedia.org/wiki/Network_tap">Wikipedia, Network tap (passive taps, 10/100, gigabit)</a></li>
<li><a href="https://www.modbus.org/file/secure/modbusprotocolspecification.pdf">Modbus Organization, Modbus Application Protocol Specification V1.1b3 (PDF)</a></li>
<li><a href="https://cars.uchicago.edu/software/epics/Modbus_over_serial_line_V1_02.pdf">Modbus Organization, Modbus over Serial Line V1.02 (PDF)</a></li>
<li><a href="https://www.hussmann.com/ns/Technical-Documents/XWi70K-EN.pdf">Dixell, XWi70K controller manual (PDF)</a></li>
<li><a href="https://assets.danfoss.com/documents/211618/AJ430138910308en-000101.pdf">Danfoss, communications between controllers and the AK-SM 800A (PDF)</a></li>
<li><a href="https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/3008/MAX3095,%2096.pdf">Maxim, MAX3095 RS-485 receiver datasheet (PDF)</a></li>
<li><a href="https://www.pusr.com/products/Serial-to-Dual-Band-WiFi-Converter.html">PUSR, USR-DR164 module datasheet, RS-485 to dual-band Wi-Fi on DIN rail</a></li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration contractor, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>A Satellite for the Store Wi-Fi</title>
      <link>https://trust-buster.com/blog/a-satellite-for-the-store-wi-fi</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/a-satellite-for-the-store-wi-fi</guid>
      <pubDate>Wed, 11 Nov 2026 11:15:00 GMT</pubDate>
      
      <description>A satellite for the store Wi-Fi: a single door to the refrigeration supervisor, open to the refrigeration engineer within the site manager&apos;s window.</description>
      <content:encoded><![CDATA[<p><strong>A satellite for the store Wi-Fi, in brief</strong></p>
<ul>
<li>Our Wi-Fi gateway, a satellite of the Cube, connects to the store&#39;s Wi-Fi. It then groups the refrigeration equipment over Ethernet behind a single door.</li>
<li>This door only opens to the refrigeration engineer within the window granted by the site manager. Every session goes into the log. Outgoing alarms are never deliberately blocked.</li>
<li>Our Wi-Fi decoy reports local attempts without striking back. It does not, however, see an attacker with stolen credentials who comes through the open tunnel.</li>
</ul>
<p><em>Fourth article in the series. The previous one, <a href="/blog/observe-without-reading">&quot;Observe without reading&quot;</a>, presents the Cube as the store&#39;s access supervisor. It describes the sentinel diodes, which observe a link without being able to speak on it. They thus read the time and duration of an intervention without opening an exchange.</em></p>
<h2 id="where-we-start-from">Where we start from</h2>
<p>The refrigeration supervisor (AK-SM at Danfoss, XWEB at Dixell, a brand of the Copeland group) is often connected in one of two ways. Either it shares the store network with the tills, reachable by anything that travels on it. Or the contractor has installed their own 4G router, <a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">as Dixell France recommends</a>. The equipment is then well isolated. But the store no longer sees what comes in. <strong>What is gained in separation is lost in visibility.</strong></p>
<p><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR</a> considers the exposure of these interfaces &quot;contrary to good practice&quot;. It therefore asks for them to be removed from the Internet and for access to be restored only through a secure VPN or a protected network. As an emergency measure, it does accept filtering the contractors&#39; IP addresses. But this solution &quot;must not be considered lasting&quot;.</p>
<p>The industrial remote maintenance gateways on the market already offer an outgoing tunnel, on-demand access and a log. Our gateway does not replace them. It adds the link to the Cube, which decides the window according to the site manager&#39;s rules. The Cube also keeps the evidence outside the store&#39;s network.</p>
<h2 id="what-a-satellite-for-the-store-wi-fi-does">What a satellite for the store Wi-Fi does</h2>
<p>The gateway is a small box placed near the supervisor. <strong>Upstream, it joins the store&#39;s Wi-Fi.</strong> <strong>Downstream, it offers an Ethernet port.</strong> The supervisor and the devices that talk to it are reconnected there. This is network segmentation at the scale of a store. Nothing reaches this segment any more without going through it.</p>
<p>Installation takes place during a visit by the refrigeration engineer, preferably during a planned intervention. They reconnect the supervisor to the box. Then they check with the site manager that the alarms still go out. No refrigeration setting is touched. It may, however, be necessary to review the supervisor&#39;s addressing, its network gateway and the server it sends alarms through. The refrigeration engineer does this or validates it.</p>
<p>On the refrigeration engineer&#39;s side, the gateway does not provide the access. The technician keeps their usual path. This is their company&#39;s VPN, the manufacturer&#39;s cloud or the retailer&#39;s bastion. This <strong>existing tunnel</strong> now ends at the box. The box lets it through or not according to the site manager&#39;s window. It logs every session and alerts outside the window. Strong authentication of the technician remains a matter for that access and its operator. Trust-Buster operates no remote access service.</p>
<h3 id="window-and-log">Window and log</h3>
<p>The refrigeration engineer&#39;s tunnel only passes during the <strong>intervention window</strong> granted in the Cube&#39;s rules. For example, from 8 am to 10 am this Tuesday. At night, the on-call person approves with a gesture on their phone. <a href="/blog/the-site-manager-rules">The last article</a> details this. IP filtering is added to the window without replacing it.</p>
<p><strong>Every session is traced</strong>: which account or which device, at what time, for how long. Within the window, no alert: the session goes into the log and the daily summary. Outside the window, on the other hand, or beyond the permitted duration, the alert is immediate. For the refrigeration company, this log is also a proof of service.</p>
<p><strong>The gateway never deliberately blocks outgoing alarms.</strong> They always pass to the known destinations. Only remote maintenance therefore depends on the window.</p>
<h2 id="a-trap-for-local-attempts">A trap for local attempts</h2>
<p>The October attacks came from the Internet. The decoy targets a local threat, less likely but cheap to watch for. It is, for example, someone trying their luck from the car park.</p>
<p>The box advertises a <strong>Wi-Fi decoy</strong> under a name specific to the site. No legitimate device has any reason to look for it (&quot;admin&quot; would attract staff phones). Nobody knows its WPA2 key. Association may succeed, but the key exchange fails. The decoy listens to one channel at a time and reports four families of events to the Cube:</p>
<ul>
<li>A device that <strong>looks for this network by its name</strong>.</li>
<li>A <strong>connection attempt</strong> whose key exchange fails.</li>
<li>A <strong>rogue access point</strong> imitating the decoy or the store, which requires knowing the list of legitimate access points.</li>
<li>A <strong>burst of deauthentication</strong>.</li>
</ul>
<p><strong>This is detection only</strong>, with no jamming or retaliation. The decoy therefore disturbs neither the tills nor the customer Wi-Fi. Its link with the Cube is, moreover, signed.</p>
<p>Since the attack, many have placed decoys on the Internet that imitate the login page of these supervisors. According to <a href="https://www.bitsight.com/blog/supermarket-refrigeration-cyberattacks-data">Bitsight</a>, they inflate the counts of exposed equipment. Ours only watches for what happens within radio range of the store.</p>
<h2 id="where-the-wi-fi-sentinel-diode-fits">Where the Wi-Fi sentinel diode fits</h2>
<p>The <strong>Wi-Fi sentinel diode</strong> listens to beacons and management frames without ever associating. Its silence, however, rests on the software. Its radio could transmit, but its firmware prevents it. The Ethernet and serial diodes, for their part, physically have nothing with which to transmit. The Wi-Fi diode, on the other hand, avoids rewiring and switches nothing off.</p>
<p>It also watches the Wi-Fi the box depends on. Jamming or a burst of disconnections would indeed cut off remote maintenance. The store&#39;s Wi-Fi must therefore protect its management frames (802.11w, mandatory under WPA3). Pure jamming, however, remains indistinguishable from a radio fault.</p>
<h2 id="if-the-gateway-fails">If the gateway fails</h2>
<p><strong>The refrigeration carries on</strong>, because the controllers drive the cabinets locally. <strong>Remote maintenance stops</strong>, which is the safe behaviour. The refrigeration engineer then reconnects the old path while the box is replaced.</p>
<p><strong>The alarms must not depend on the gateway alone.</strong> Otherwise, a fault in the box would silence them. We therefore recommend keeping a separate path, which often exists. It may be a contact alarm output, a transmitter or a contractor&#39;s router reserved for alarms. The Cube notices that the satellite for the store Wi-Fi has gone quiet. It reports it at once over its own cellular link, and the site manager&#39;s phone can also relay the alert.</p>
<h2 id="the-questions-to-ask-before-installation">The questions to ask before installation</h2>
<p>Is the supervisor on Ethernet? Does it talk to other devices that must be moved with it? Is the Wi-Fi shared with customers or managed by the retailer? In that case it is often the retailer that must be called. Does the supervisor talk constantly to an online service? That flow will then have to be authorised separately, outside the window. Your refrigeration engineer usually knows the answers. The installation is, in any case, decided with them.</p>
<h2 id="what-the-satellite-for-the-store-wi-fi-proves-and-what-it-do">What the satellite for the store Wi-Fi proves, and what it does not</h2>
<table>
<thead>
<tr>
<th>Mechanism</th>
<th>Proves</th>
<th>Does not prove</th>
<th>Who answers for it</th>
</tr>
</thead>
<tbody><tr>
<td>Segment and tunnel</td>
<td>No planned network path reaches the supervisor outside the gateway</td>
<td>The legitimacy of the session</td>
<td>The site manager, and us for the gateway</td>
</tr>
<tr>
<td>Window and log</td>
<td>When and for how long</td>
<td>What was done</td>
<td>The site manager</td>
</tr>
<tr>
<td>IP filtering</td>
<td>A known address</td>
<td>The right person</td>
<td>The contractor</td>
</tr>
<tr>
<td>Wi-Fi decoy</td>
<td>A local attempt</td>
<td>The absence of others</td>
<td>The site manager, alerted</td>
</tr>
</tbody></table>
<p>The first line assumes that the old wiring, a second Ethernet port, an old 4G router or a built-in modem has been removed.</p>
<p><strong>The trap does not see an attacker with stolen credentials who comes through the open tunnel.</strong> This is the main limit, because such an attacker looks like the expected technician. What remains are the signals of the previous article: a longer session or an unusual time. If the traffic is read, a setpoint outside the site&#39;s habits is added to them.</p>
<p><strong>In the next article</strong>, <a href="/blog/listen-without-being-able-to-speak">&quot;Listen without being able to speak&quot;</a>, we describe the Ethernet and serial sentinel diodes. They read the traffic without being able to send a command.</p>
<p><em>Do you operate or maintain remotely maintained installations? Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>. We will show you on a demonstration bench how these signals can be observed.</em></p>
<h2 id="sources">Sources</h2>
<ul>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a></li>
<li><a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">CTOL, Dixell France guidance and end of support for XWEB EVO</a></li>
<li><a href="https://www.bitsight.com/blog/supermarket-refrigeration-cyberattacks-data">Bitsight, data on attacks against supermarket refrigeration supervisors</a></li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration engineer, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Observe Without Reading: The Legitimacy of Remote Access</title>
      <link>https://trust-buster.com/blog/observe-without-reading</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/observe-without-reading</guid>
      <pubDate>Wed, 11 Nov 2026 11:10:00 GMT</pubDate>
      
      <description>Observing without reading the traffic: the time, duration and origin of a remote access are often enough to spot unusual remote maintenance.</description>
      <content:encoded><![CDATA[<p><strong>Observe without reading, in brief</strong></p>
<ul>
<li>A remote intervention leaves traces. We observe them without reading the traffic: who, from where, when, for how long, how often, on how many sites.</li>
<li>These traces are only meaningful when compared with the site&#39;s usual behaviour. Only the value of a setpoint requires reading what passes through.</li>
<li>We describe the Cube, the site&#39;s access supervisor, then our <strong>sentinel diodes</strong>. These satellites observe the Wi-Fi, the Ethernet or the serial link without being able to write to it.</li>
</ul>
<p><em>Third article in the series. The previous one, <a href="/blog/why-the-fridges-stayed-open">&quot;Why the fridges stayed open&quot;</a>, listed seven combined weaknesses and why every remedy runs into the realities of the field. It concluded that control will remain possible, but that nothing called its legitimacy into question.</em></p>
<h2 id="the-site-manager-s-question">The site manager&#39;s question</h2>
<p>Take the manager of a site, whether it is a shop, a workshop, a factory or a warehouse. Last night, someone connected to the refrigeration supervisor. The manager has no way of knowing whether it was their refrigeration engineer. To help them, we propose to <strong>observe without reading</strong> what passes through.</p>
<p>After the attack, one often read that &quot;the air gap is a myth, move to Zero Trust&quot;. Remote maintenance has indeed broken the isolation (the air gap) of these installations. But Zero Trust, where every access is verified, remains the name of a category. It does not say who will pay for it in hundreds of shops.</p>
<p><a href="https://www.ic3.gov/CSA/2026/261105.pdf">The US agencies</a> are more modest. First, they isolate industrial equipment on a dedicated virtual network (VLAN). Then they restrict access to authorised addresses. They ask for the logs to be reviewed. They also recommend to &quot;define and maintain reasonable setpoint limits&quot; and to &quot;establish alerting for deviations&quot;. We therefore entrust this work to a box that does not depend on the supervisor it watches.</p>
<h2 id="refrigeration-remote-maintenance-the-signs-of-a-legitimate-i">Refrigeration remote maintenance: the signs of a legitimate intervention</h2>
<p>A refrigeration engineer who works remotely has habits, even unwritten ones. They connect from their branch office, during the day unless there is a breakdown, for a few minutes. Above all, they rarely target dozens of sites in the same hour. These habits form the site&#39;s <strong>baseline behaviour</strong>. It then makes it possible to judge an OT remote access without reading its content.</p>
<table>
<thead>
<tr>
<th>Question</th>
<th>What is observed</th>
<th>Must we read?</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Who</strong></td>
<td>The device or the account, if it can be identified</td>
<td>No</td>
</tr>
<tr>
<td><strong>From where</strong></td>
<td>Behind a tunnel, only the address of the router, the VPN concentrator or the cloud</td>
<td>No</td>
</tr>
<tr>
<td><strong>When</strong></td>
<td>The time, compared with the usual hours</td>
<td>No</td>
</tr>
<tr>
<td><strong>How much</strong></td>
<td>The duration and the volume exchanged</td>
<td>No</td>
</tr>
<tr>
<td><strong>How often</strong></td>
<td>The number of sessions per week</td>
<td>No</td>
</tr>
<tr>
<td><strong>With what value</strong></td>
<td>The setpoint written, for example a maximum defrost duration</td>
<td>Yes</td>
</tr>
<tr>
<td><strong>On how many sites</strong></td>
<td>The same action, the same night, elsewhere</td>
<td>Several sites must be seen</td>
</tr>
</tbody></table>
<p>The four signals of the previous article (time, value, duration, simultaneity) are found again in these seven questions. To see simultaneity, the administration server pools the alerts of several sites.</p>
<p>The baseline behaviour is learnt over at least one full week. It then lacks the monthly or seasonal interventions. As long as it remains thin, rules written with the refrigeration engineer are worth more. An anomaly already under way would, moreover, be learnt as normal. It is therefore better to check the configuration first, in particular the defrost ranges, as <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR</a> asks.</p>
<h2 id="observe-without-reading-then-read-only-what-matters">Observe without reading, then read only what matters</h2>
<p>Observing without reading is not an affectation. Exchanges with the supervisor are increasingly encrypted, though not always on older models. Yet the metadata remains visible. It is enough to spot many unusual behaviours, without touching the refrigeration settings.</p>
<p>That leaves the value. Temperatures <a href="https://www.it-connect.fr/cyberattaque-supermarches-froid-danfoss/">pushed up to 50 °C on some sites</a> cannot be guessed from the duration of a session. Part of the setpoints can be read in clear on the serial link. This link connects the supervisor to its controllers, often over Modbus RTU. Only part of them, because defrost schedules and coordinated functions live in the supervisor. On the bus, they only pass as start commands. No public source names Modbus as the way the attacks got in. The point is only to read there what the supervisor has ordered.</p>
<h2 id="the-sentinel-diodes">The sentinel diodes</h2>
<p>The Cube is a small box placed in the plant room. It sends its logs and alerts over its own cellular link, and the site manager&#39;s phone can act as a relay. Here, it serves as the site&#39;s <strong>access supervisor</strong>, not to be confused with the refrigeration supervisor. It does not regulate the cold, but watches who accesses those who regulate it. It relies on satellites called <strong>sentinel diodes</strong>, because they observe without being able to write.</p>
<p>The word diode is an image. These are not certified data diodes: they do not claim that level of assurance. Downstream, the Cube talks to an administration server, the operator&#39;s or ours depending on the contract. This link carries only signed rules and logs, never a command to the refrigeration.</p>
<h3 id="three-sentinels-from-wi-fi-to-the-serial-bus">Three sentinels, from Wi-Fi to the serial bus</h3>
<p><strong>The Wi-Fi sentinel</strong> spots the devices that associate, without decrypting the traffic. It listens to 2.4 GHz, channel by channel, but sees neither 5 GHz nor a wired supervisor. A radio, however, can transmit: its silence rests on the software. We therefore place it in the family for its role.</p>
<p><strong>The Ethernet diode</strong> sits as a passive tap on the supervisor&#39;s cable. It sees the sessions, their origins and their durations, without being reachable on the network.</p>
<p><strong>The serial diode</strong> connects to the RS-485 link when the controllers speak Modbus RTU. Its circuit has no transmit stage. A one-way optocoupler links it to the Cube. The RS-232 of some equipment, however, requires a different receiver.</p>
<p>For Ethernet and serial, on the other hand, the one-way direction rests on the hardware. No component of the circuit can transmit. The diodes switch nothing off. If the site manager has provided for it, another satellite suspends incoming remote access. It never targets the refrigeration or the outgoing alarms. A controlled plug, moreover, only switches off running equipment if it appears on the site manager&#39;s list, that of the <a href="/blog/power-off-is-not-an-emergency-stop">equipment that can be cut while running</a>.</p>
<h2 id="which-satellite-for-which-existing-setup">Which satellite for which existing setup</h2>
<p>The Cube adapts to the existing setup, which your refrigeration engineer can describe.</p>
<ul>
<li><strong>On the shop&#39;s Wi-Fi</strong>: the Wi-Fi gateway of article 4, with the Wi-Fi sentinel as a complement.</li>
<li><strong>Over Ethernet</strong>: the Ethernet diode, which does not see the setpoint when the interface is encrypted.</li>
<li><strong>Controllers on Modbus RTU</strong>: the serial diode, which reads the setpoints but without knowing who gave the order.</li>
<li><strong>The contractor&#39;s 4G router</strong>: an Ethernet diode behind the router. With the contractor&#39;s written agreement, a controlled plug is also added on its power supply.</li>
<li><strong>Manufacturer cloud</strong>: an Ethernet diode on the outgoing side, the technician&#39;s identity remaining known to the manufacturer alone.</li>
</ul>
<p>A word of caution: switching off a 4G router that also carries the alarms switches them off with it. They therefore first need a separate path. On large sites, bastions and industrial probes already do this work, with a dedicated team. We are aiming at the site that has neither.</p>
<h2 id="what-observation-proves-and-who-answers-for-it">What observation proves, and who answers for it</h2>
<p><strong>Observation proves an anomaly of behaviour, not an intention.</strong> A session at three in the morning may be a genuine breakdown (false positive). But a session at ten o&#39;clock, from the usual address, may come from an attacker (false negative).</p>
<p>An announced intervention within a granted window triggers no alert. It goes into the log and the daily summary. Outside the window, on the other hand, or beyond the permitted span, the alert is immediate. The false positive then costs a call to the refrigeration engineer, then a gesture to acknowledge it on the phone. The false negative is more serious. An attacker who imitates the baseline behaviour escapes the metadata. Observing without reading therefore has its limits, and reading the setpoint reduces this risk without removing it.</p>
<p>Who answers for it? The refrigeration engineer remains responsible for their interventions, and the site manager for their rules. Our satellites keep two promises: being unable to write, and faithfully logging what they have seen. Our commitments on these two points are therefore set out in the contract.</p>
<p>The next article, <a href="/blog/a-satellite-for-the-store-wi-fi">&quot;A satellite for the store Wi-Fi&quot;</a>, details this frequent case.</p>
<p><em>Do you know how your supervisor is connected, or would you like to know? Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>.</em></p>
<h2 id="sources">Sources</h2>
<ul>
<li><a href="https://www.ic3.gov/CSA/2026/261105.pdf">CISA, FBI, HHS, USDA and DoT, &quot;Hardening Industrial Refrigeration and Cold Chain Monitoring Systems&quot; fact sheet of 5 November 2026 (PDF)</a></li>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a></li>
<li><a href="https://www.it-connect.fr/cyberattaque-supermarches-froid-danfoss/">IT-Connect, the supermarket cyberattack and the 50 °C defrost setpoint</a></li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration engineer, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Why the Fridges Stayed Open</title>
      <link>https://trust-buster.com/blog/why-the-fridges-stayed-open</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/why-the-fridges-stayed-open</guid>
      <pubDate>Wed, 11 Nov 2026 11:05:00 GMT</pubDate>
      
      <description>Exposed access, default password, end-of-life hardware: why the fridges stayed open, and what remote maintenance makes difficult to fix.</description>
      <content:encoded><![CDATA[<p><strong>Why the fridges stayed open, in brief</strong></p>
<ul>
<li>Seven weaknesses added up. Four are well known to everyone, and the other three explain why the first four lasted.</li>
<li>Every good practice runs up against a real constraint, because remote maintenance is a business need. A VPN says nothing about what is done inside the tunnel, and encryption does not judge intent.</li>
<li>Exposure and passwords opened the door. What was missing next was someone to question the <strong>legitimacy</strong> of the commands being sent.</li>
</ul>
<p><em>Second article in the series, after <a href="/blog/hacked-fridges-defrosting-from-afar">“Hacked fridges: defrosting from afar”</a>, which recalls the facts. In short, on the night of 11 to 12 October 2026, attackers used interfaces exposed on the Internet to reach the refrigeration supervisors of some 700 retail sites, where they prolonged the defrost cycles and drove temperatures as high as 50 °C on some sites.</em></p>
<p><em>We did not live through these incidents on the ground, but we have known similar ones. What follows is therefore our analysis, informed by that field experience.</em></p>
<h2 id="seven-weaknesses">Seven weaknesses</h2>
<table>
<thead>
<tr>
<th>Weakness</th>
<th>Finding</th>
<th>Why it lasts</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Exposed administration interface</strong></td>
<td>Judged <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">contrary to good practice by CERT-FR</a>. Censys counted <a href="https://www.analyticsinsight.net/news/us-agencies-warn-of-cyberattacks-on-industrial-refrigeration-systems">more than 3,000 web interfaces of Danfoss supervisors visible worldwide on 9 October</a>.</td>
<td>It is the simplest way to offer remote maintenance.</td>
</tr>
<tr>
<td><strong>Weak or default passwords</strong></td>
<td><a href="https://www.publicsenat.fr/actualites/politique/des-defauts-de-securisation-elementaire-et-des-mots-de-passe-faibles-ont-permis-la-cyberattaque-contre-les-frigos-de-supermarches">Cited by ANSSI</a>. Some supervisors had an <a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">“Admin” account shipped with the password “Admin”</a>.</td>
<td>The password is also the shared secret of a trade.</td>
</tr>
<tr>
<td><strong>End-of-life equipment</strong></td>
<td><a href="https://hard2bit.com/en/blog/french-supermarket-cyberattack-refrigeration-danfoss/">AK-SM 800 without support since 2021, AK-SC 355 since 2016</a>.</td>
<td>A supervisor often outlives its manufacturer support by many years.</td>
</tr>
<tr>
<td><strong>Known vulnerabilities left unpatched</strong></td>
<td>Not exploited this time according to ANSSI, but <a href="https://refindustry.com/fr/news/danfoss-copeland-et-trane-corrigent-des-failles-de-securite-des-controleurs/">published and fixed</a>, without the fix being applied everywhere.</td>
<td>Tracking patches for this hardware is often nobody&#39;s job.</td>
</tr>
</tbody></table>
<p>None of these weaknesses is new, for that matter.</p>
<h3 id="why-they-lasted">Why they lasted</h3>
<table>
<thead>
<tr>
<th>Weakness</th>
<th>Finding</th>
<th>Why it lasts</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Legitimate command misused</strong></td>
<td>Defrosting is a normal function. Prolonged, it drove temperatures to 50 °C on some sites.</td>
<td>US agencies now call for <a href="https://www.ic3.gov/CSA/2026/261105.pdf">setpoints to be bounded within reasonable ranges</a>.</td>
</tr>
<tr>
<td><strong>No independent detection</strong></td>
<td>Where only the supervisor raised the alarm, nothing went off. <a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">The EFS</a>, which had an on-call rota, is the exception.</td>
<td>The compromised supervisor is also the one that produces the alarms.</td>
</tr>
<tr>
<td><strong>Scattered responsibilities</strong></td>
<td><a href="https://www.frenchweb.fr/frigos-pirates-les-installations-locales-angle-mort-de-la-cybersecurite-des-enseignes/900000015">Management shared</a> between the retail chain, the store, the refrigeration engineer and the manufacturer.</td>
<td>Nobody therefore holds the complete inventory.</td>
</tr>
</tbody></table>
<p><strong>According to the published information, the attacker did not have to break any protection on the equipment.</strong> They sent a command that a technician could have sent. What was abnormal was the time, the value, the duration, and hundreds of sites receiving the same command on the same night.</p>
<h2 id="remote-maintenance-and-remote-access-why-every-remedy-runs-u">Remote maintenance and remote access: why every remedy runs up against reality</h2>
<p><strong>Working remotely is a business need.</strong> A refrigeration engineer monitors dozens, sometimes hundreds of installations at night. <a href="https://info.agriculture.gouv.fr/boagri/instruction-N2006-8247/telechargement">French regulations</a> require the temperature of frozen food warehouses to be recorded and those readings to be kept. Moreover, <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR asks for the insurance conditions to be checked</a> before cutting off the flow of data to the refrigeration engineer. <strong>The remote link is therefore part of the security of the cold chain.</strong></p>
<p><strong>A VPN says nothing about what is done inside the tunnel.</strong> <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR</a> rightly recommends going through a VPN, but access often relies on accounts shared between technicians, and VPN appliances are themselves <a href="https://www.it-connect.fr/attaque-massive-brute-force-cible-firewalls-et-vpn-avec-2-8-millions-adresses-ip/">the target of brute-force campaigns</a>. As for the contractors&#39; 4G routers, <a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">which one manufacturer recommends</a>, they do isolate the equipment, but they make it invisible to the store.</p>
<p><strong>Encryption protects the content, not the intent.</strong> It is indispensable, yet an attacker holding a valid credential sends encrypted commands, and the defrost setpoint would have been just as destructive inside a flawless tunnel.</p>
<h3 id="a-matter-of-organisation">A matter of organisation</h3>
<p><strong>The default password is an organisational problem.</strong> On one site, changing it certainly takes a minute, but across hundreds of sites, with rotating on-call technicians and subcontractors, it is a project. <a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">The guidance published since the attack</a> calls for named accounts and expiry dates for those who intervene, which is identity management, still poorly tooled in this trade. We discussed this in <a href="/iam-electric">An IAM for your electrical equipment</a>.</p>
<p><strong>In most stores, nobody watches the administration channel.</strong> The supervisor both controls and alerts, so once compromised it can bring down both. At the EFS, an on-call technician and the alarms are said to have spotted the anomalies in time, and no blood products were lost.</p>
<p><strong>And responsibility remains diffuse.</strong> The manufacturer sells a product, the installer connects it, the refrigeration engineer maintains it, the store pays for it and the retail chain bears the consequences. <a href="https://www.techniques-ingenieur.fr/actualite/breves-afp/la-cyberattaque-des-frigos-de-supermarches-en-france-liee-a-des-defauts-de-securisation-elementaire-selon-lanssi-165159/">ANSSI does not have the means to compel</a> a supermarket, and the transposition of NIS2 <a href="https://www.solutions-numeriques.com/?p=251486">is still under discussion</a>. For food, moreover, <a href="https://streamlex.eu/annexes/nis2-en-annex-ii/">the directive covers wholesale distribution and industrial production</a>, not necessarily the store. <strong>Nobody is in charge of the security of the administration channel. In practice, however, it is the site that answers for the lost goods.</strong></p>
<blockquote>
<p><strong>Update of 11 November 2026.</strong> The Resilience bill, which transposes NIS2, was <a href="https://www.assemblee-nationale.fr/dyn/actualites-accueil-hub/examen-du-projet-de-loi-relatif-a-la-resilience-des-infrastructures-critiques-et-au-renforcement-de-la-cybersecurite">debated in plenary at the French National Assembly on 9 and 10 November</a>, with a formal vote announced for 17 November. It has yet to complete its passage, and the question of scope for food retail remains open.</p>
</blockquote>
<h2 id="four-questions-to-ask-right-now">Four questions to ask right now</h2>
<ul>
<li>To your refrigeration engineer: can my supervisor be reached from the Internet? By which path, including a 4G maintenance router?</li>
<li>Again to your refrigeration engineer: who has an account on it? Has the original password been changed?</li>
<li>To your refrigeration engineer: does my model still receive patches from the manufacturer?</li>
<li>To your insurer: does my contract require temperatures to be forwarded, and what does it provide for in the event of remote sabotage?</li>
</ul>
<blockquote>
<p><strong>Update of 11 November 2026.</strong> According to brokers quoted by <a href="https://www.insurancejournal.com/news/international/2026/10/15/889037.htm">Reuters</a>, cyber policies do not generally cover physical losses, and the war exclusion could complicate claims when a state is designated. In France, compensation for losses caused by an attack on a computer system also requires a complaint to be filed no later than 72 hours after its discovery (<a href="https://www.legifrance.gouv.fr/codes/article_lc/LEGIARTI000047048152">Article L12-10-1 of the French Insurance Code</a>).</p>
</blockquote>
<h2 id="if-the-fridges-stayed-open-what-we-take-from-it">If the fridges stayed open, what we take from it</h2>
<p>Let us go back over the night of 11 to 12 October as an experienced refrigeration engineer would. <strong>Four signals stand out: the time, the value, the duration and the simultaneity.</strong></p>
<p>The time and the duration of a session can be seen without reading anything of the exchanges. The value, on the other hand, requires reading the setpoint, and the simultaneity can only be seen by looking at several sites. None is enough on its own, since a night-time defrost is commonplace, but each was enough to raise the question.</p>
<p>The fridges stayed open first because their supervisors could be reached from the Internet, with weak passwords or passwords that were never changed. That is therefore the first thing to fix. But even once it is fixed, remote control will remain possible, since remote maintenance requires it.</p>
<p>What was missing next was anything to question its <strong>legitimacy</strong>. Who is sending it, from where, when, and with what amplitude? Does it look like what this site usually does? This question can be handled without reading the exchanges, without replacing the installed equipment and without cutting the refrigeration engineer&#39;s link. It can also do without the very supervisor that needs protecting.</p>
<h3 id="a-last-line-of-defence">A last line of defence</h3>
<p>The site bears the consequences, while its manager generally has neither a security team nor a dedicated budget, and that is as true of a store as of a workshop, a factory or a warehouse. The site therefore needs a last line of defence, which alerts the manager and suspends incoming remote access according to rules the manager has set, and which reopens access just as simply when the refrigeration engineer needs it. Suspending incoming access touches neither the cooling nor the forwarding of alarms, provided those alarms have a separate path, since the controllers carry on working on site.</p>
<p><strong>In the rest of the series</strong>, we describe our answer, the Trust-Buster Cube and its satellites. The Cube is a unit installed on the site, and its satellites plug into the existing installation without touching the refrigeration settings. They replace as little as possible, because the Cube adapts to what is already there, not the other way round. Each time, we say what the mechanism proves, what it does not prove, and who answers for it:</p>
<ol start="3">
<li><a href="/blog/observe-without-reading"><em>Observe without reading</em></a>: the signals of legitimacy of a remote intervention.</li>
<li><a href="/blog/a-satellite-for-the-store-wi-fi"><em>A satellite for the store Wi-Fi</em></a>: a single controlled door to the refrigeration equipment.</li>
<li><a href="/blog/listen-without-being-able-to-speak"><em>Listen without being able to speak</em></a>: sentinel diodes and what they change for the warranty. They read the exchanges without being able to send a command.</li>
<li><a href="/blog/the-site-manager-rules"><em>The site manager&#39;s rules</em></a>: alert, suspend, revoke, and keep the proof.</li>
</ol>
<p><em>Do you operate or maintain remotely maintained installations? Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>. We will show you on a demonstration bench how these signals can be observed.</em></p>
<h2 id="sources">Sources</h2>
<p>Authorities and official texts:</p>
<ul>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a>.</li>
<li><a href="https://info.agriculture.gouv.fr/boagri/instruction-N2006-8247/telechargement">French Ministry of Agriculture, recording the temperatures of frozen food</a>.</li>
<li><a href="https://www.ic3.gov/CSA/2026/261105.pdf">FBI / CISA, advisory of 5 November 2026 (PDF)</a>.</li>
<li><a href="https://streamlex.eu/annexes/nis2-en-annex-ii/">NIS2 Directive, Annex II</a>.</li>
<li><a href="https://www.assemblee-nationale.fr/dyn/actualites-accueil-hub/examen-du-projet-de-loi-relatif-a-la-resilience-des-infrastructures-critiques-et-au-renforcement-de-la-cybersecurite">French National Assembly, debate on the Resilience bill (9 and 10 November 2026)</a>.</li>
<li><a href="https://www.legifrance.gouv.fr/codes/article_lc/LEGIARTI000047048152">French Insurance Code, Article L12-10-1</a>.</li>
</ul>
<p>Press and analyses:</p>
<ul>
<li><a href="https://www.publicsenat.fr/actualites/politique/des-defauts-de-securisation-elementaire-et-des-mots-de-passe-faibles-ont-permis-la-cyberattaque-contre-les-frigos-de-supermarches">Public Sénat, hearing of the Director General of ANSSI (21 October 2026)</a>.</li>
<li><a href="https://www.techniques-ingenieur.fr/actualite/breves-afp/la-cyberattaque-des-frigos-de-supermarches-en-france-liee-a-des-defauts-de-securisation-elementaire-selon-lanssi-165159/">AFP via Techniques de l&#39;Ingénieur, the limits of ANSSI&#39;s powers</a>.</li>
<li><a href="https://www.analyticsinsight.net/news/us-agencies-warn-of-cyberattacks-on-industrial-refrigeration-systems">Analytics Insight, exposure according to Censys</a>.</li>
<li><a href="https://www.ctol.digital/news/refrigeration-controller-attacks-cert-fr-xweb-evo-end-of-service/">CTOL, Dixell France guidance and end of support for XWEB EVO</a>.</li>
<li><a href="https://hard2bit.com/en/blog/french-supermarket-cyberattack-refrigeration-danfoss/">Hard2bit, affected Danfoss models and end of life</a>.</li>
<li><a href="https://refindustry.com/fr/news/danfoss-copeland-et-trane-corrigent-des-failles-de-securite-des-controleurs/">Refindustry, Danfoss and Copeland vulnerabilities fixed</a>.</li>
<li><a href="https://www.frenchweb.fr/frigos-pirates-les-installations-locales-angle-mort-de-la-cybersecurite-des-enseignes/900000015">FrenchWeb, local installations, the retail chains&#39; blind spot</a>.</li>
<li><a href="https://www.it-connect.fr/attaque-massive-brute-force-cible-firewalls-et-vpn-avec-2-8-millions-adresses-ip/">IT-Connect, brute force against firewalls and VPNs</a>.</li>
<li><a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">The EFS, quoted by Le Quotidien du Médecin</a>.</li>
<li><a href="https://www.solutions-numeriques.com/?p=251486">Solutions Numériques, the Resilience bill back at the National Assembly</a>.</li>
<li><a href="https://www.insurancejournal.com/news/international/2026/10/15/889037.htm">Reuters via Insurance Journal, cyber insurance and physical losses (15 October 2026)</a>.</li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration engineer, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Hacked Fridges: Defrosting from Afar</title>
      <link>https://trust-buster.com/blog/hacked-fridges-defrosting-from-afar</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/hacked-fridges-defrosting-from-afar</guid>
      <pubDate>Wed, 11 Nov 2026 11:00:00 GMT</pubDate>
      
      <description>Hacked fridges: in October 2026, 700 supermarkets saw their refrigeration supervisors forced into defrost. The facts, and what the debate confused.</description>
      <content:encoded><![CDATA[<p><strong>Hacked fridges, in brief</strong></p>
<ul>
<li>The cyberattack on supermarket fridges took place on the night of 11 to 12 October 2026. It targeted the refrigeration supervisors of some 700 French sites and prolonged their defrost cycles, so that temperatures climbed as high as 50 °C on some sites. Three weeks later, the same scenario played out again in Europe and at the French blood service, the Établissement français du sang.</li>
<li>The attack needed no new vulnerability. It took advantage of interfaces exposed on the Internet, of weak or default passwords and of end-of-life equipment.</li>
<li>Yet the public debate often mixed up three episodes and two attributions. It also confused two worlds, business IT and industrial IT.</li>
</ul>
<p><em>We did not live through these incidents on the ground, but we have known similar ones in contexts every bit as sensitive.</em></p>
<p><em>Now that the commotion has died down, here is our analysis of the facts and the problems behind the hacked fridges, so as to offer you an informed opinion grounded in our understanding and in our own field experience.</em></p>
<p><em>This first article in a series is written for site managers, whether of a store, a workshop, a factory or a warehouse. It is also meant for refrigeration engineers, maintenance contractors and anyone who owns connected industrial equipment. It recalls the facts, then sorts through what has been said about them.</em></p>
<h2 id="hacked-fridges-the-facts">Hacked fridges: the facts</h2>
<h3 id="the-night-of-11-to-12-october-in-france">The night of 11 to 12 October, in France</h3>
<p>The attack did not target the refrigerators themselves but the <strong>supervisors</strong> that control them. These devices centralise the controllers of the display cabinets and cold rooms, record temperatures and forward alarms, and they also manage the defrost cycles, while the refrigeration engineer connects to them remotely. That night, the attacker pushed their defrost parameters to extreme values, and temperatures then rose <a href="https://fr.euronews.com/2026/11/09/la-france-accuse-liran-de-cyberattaques-contre-les-frigos-de-supermarches">as high as 50 °C in some cases</a>. According to <a href="https://cyber.gouv.fr/actualites/campagnes-de-d%C3%A9stabilisation-et-sabotage-men%C3%A9es-%C3%A0-lencontre-des-int%C3%A9r%C3%AAts-fran%C3%A7ais-par-la-r%C3%A9publique-islamique-diran/">the assessment published by ANSSI</a>, the French national cybersecurity agency, the attack targeted some <strong>700 sites</strong> and caused at least 15 outbreaks of fire. For a site, this means goods to be thrown away and premises to be inspected, sometimes with an insurance claim on top.</p>
<p><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">The CERT-FR alert</a> concerns Danfoss AK-SM and AK-SC supervisors, as well as the XWEB units from Dixell, a brand of the Copeland group, and it identifies one main cause, an administration interface exposed on the Internet. <a href="https://hard2bit.com/en/blog/french-supermarket-cyberattack-refrigeration-danfoss/">According to Danfoss</a>, the affected units were mostly end-of-life models, which the manufacturer no longer updates. Before the French Senate on 21 October, the head of ANSSI spoke of <a href="https://www.publicsenat.fr/actualites/politique/des-defauts-de-securisation-elementaire-et-des-mots-de-passe-faibles-ont-permis-la-cyberattaque-contre-les-frigos-de-supermarches">“basic security failings”</a>, by which he meant direct exposure, unprotected administration and weak passwords. He also stated that the attack had not exploited <a href="https://www.techniques-ingenieur.fr/actualite/breves-afp/la-cyberattaque-des-frigos-de-supermarches-en-france-liee-a-des-defauts-de-securisation-elementaire-selon-lanssi-165159/">the vulnerabilities published by researchers</a>.</p>
<h3 id="november-a-second-wave-and-an-attribution">November, a second wave and an attribution</h3>
<p><strong>The night of 1 to 2 November, in Europe.</strong> The same scenario struck <a href="https://www.helpnetsecurity.com/2026/11/02/european-supermarkets-refrigeration-units-cyberattacks/">stores in seven countries</a>. In the Netherlands, the targeted systems were not on a protected network, as a Danfoss spokesperson <a href="https://xenospectrum.com/en/danfoss-freezer-cyberattack-second-wave/">quoted by Xenospectrum</a> states, adding that <strong>nobody had ever changed their default passwords</strong>. Two pro-Russian groups <a href="https://cybernews.com/security/russian-hackers-hacked-european-refrigeration-systems/">later claimed this wave</a>, a claim that nobody has verified to date.</p>
<p><strong>The same night, the French blood service.</strong> An attacker compromised the refrigeration contractor of <a href="https://selectra.info/telecom/actualites/marche/efs-cyberattaque-chambres-froides-sang">three sites of the EFS</a> and then altered temperature and defrost setpoints. An on-call technician and the alarms are, however, said to have spotted the anomalies in time, and <a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">the EFS states that it lost no blood products</a>.</p>
<p><strong>9 November, the attribution.</strong> France <a href="https://fr.euronews.com/2026/11/09/la-france-accuse-liran-de-cyberattaques-contre-les-frigos-de-supermarches">attributes the October episode</a> to the Shahid Kaveh unit of Iran&#39;s Revolutionary Guards, on the basis of a body of evidence. <a href="https://next.ink/266516/cyberattaque-contre-les-frigos-de-700-supermarches-pourquoi-la-france-accuse-liran/">The investigation</a> also notes, from the same actor, large-scale reconnaissance of other industrial automation systems using Modbus TCP, a protocol that provides neither authentication nor encryption, but the investigation does not name it as the way in. This attribution covers neither the November wave nor the EFS.</p>
<h3 id="the-c4-report">The C4 report</h3>
<blockquote>
<p><strong>Update of 11 November 2026.</strong> The <a href="https://cert.cyber.gouv.fr/uploads/CERTFR-2026-CTI-009.pdf">report of the French Cyber Crisis Coordination Centre</a>, published on 9 November by CERT-FR, details this body of evidence: attack infrastructure used during the same period for reconnaissance, use of the FOFA search tool, coverage of the attack by a media outlet as early as 13 October, and traces of Farsi. It mentions other victims in the food industry, including several small butcher&#39;s shops, a slaughterhouse and cooperative plants, as well as the concurrent targeting of equipment in small hydroelectric installations, with effects described as moderate. It does not say how the refrigeration controllers were reached, and it deals neither with the November wave nor with the EFS.</p>
<p>The same day, the Paris prosecutor&#39;s office <a href="https://www.franceinfo.fr/internet/securite-sur-internet/cyberattaques/cyberattaque-contre-des-frigos-de-supermarches-le-parquet-de-paris-ouvre-une-enquete-visant-une-unite-des-gardiens-de-la-revolution-iraniens_8274806.html">opened a judicial investigation</a> targeting the head of the unit and his two deputies. The French Ministry for Foreign Affairs <a href="https://us.diplomatie.gouv.fr/en/declaration-attribution-iran-cyber-attacks-against-france">summoned the Iranian ambassador</a> and announced that it would propose sanctions to its European partners, and Iran <a href="https://english.news.cn/20261111/9c874659cb83487db27ad65234968067/c.html">rejected these accusations</a> on 10 November.</p>
</blockquote>
<h3 id="what-remains-unknown">What remains unknown</h3>
<p>We still do not know the path the attacker took on each site. They may have used a default or guessed password, while a reused technician account, a 4G link or a VPN also remain possible. <a href="https://www.analyticsinsight.net/news/us-agencies-warn-of-cyberattacks-on-industrial-refrigeration-systems">According to Censys</a>, the Danfoss StoreView interfaces visible from the Internet fell from 3,004 on 9 October to 708 on 5 November, and the XWEB units from 3,708 to 2,420. However, according to <a href="https://www.bitsight.com/blog/supermarket-refrigeration-cyberattacks-data">Bitsight</a>, decoys set up since the attack inflate these figures. Is yours among them? Ask your refrigeration engineer.</p>
<h2 id="what-the-general-public-took-away">What the general public took away</h2>
<p>Three readings of the hacked fridges then dominated. Each contains some truth, but each shifts the subject a little.</p>
<p><strong>“Connected fridges are getting hacked.”</strong> The word is misleading, since it was not the refrigerated cabinet that could be reached but the supervisor of the installation, a professional device managed by a third party. The advice later given to households is not wrong, but it is about a different object. This confusion nonetheless has a cost, because it turns a subcontracting problem into a matter of individual hygiene.</p>
<p><strong>“Why on earth connect a fridge to the Internet?”</strong> It is the right question, and it has an answer, remote maintenance. A refrigeration engineer monitors dozens, sometimes hundreds of installations at night, and recording the temperatures of frozen food is mandatory.</p>
<p><strong>“This is hybrid warfare.”</strong> This narrative came before any evidence, and the official attribution, a month later, did not confirm the first hypotheses. <strong>For a site manager, however, the question of who did it is secondary.</strong> A military unit and an amateur would no doubt have come through the same door. Their intentions and their persistence certainly differ, but the attribution changes the diplomatic response, not the vulnerability.</p>
<blockquote>
<p><strong>Update of 11 November 2026.</strong> Although the designated perpetrator is not the one of the first hypotheses, the official texts do describe a state operation: the C4 speaks of attacks “aimed at destabilisation”, and the Ministry for Foreign Affairs sees in them a sabotage campaign on French soil. The attribution changes the diplomatic response, not the way in.</p>
</blockquote>
<h2 id="what-the-professionals-said">What the professionals said</h2>
<p>Specialists spoke a great deal, often accurately. Two shortcuts nonetheless came back again and again.</p>
<h3 id="the-solutions-that-seemed-obvious">The solutions that seemed obvious</h3>
<p><strong>“All it took was changing the password and updating.”</strong> Partly true, but most of the affected units were end of life, so no update was possible. <a href="https://selectra.info/telecom/actualites/marche/frigos-supermarches-pirates-vague-europe">A Belgian installer</a> points out that the installation then has to be physically modified.</p>
<p><strong>“They should never have been connected to the Internet.”</strong> That is indeed <a href="https://www.24matins.fr/frigos-pirates-en-supermarche-lanssi-denonce-des-erreurs-basiques-1422863">the first recommendation of the 2014 ANSSI guide</a>. But <a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR itself</a> asks for the insurance contract to be reread before cutting off the flow of temperatures to the refrigeration engineer. The link exists because it is useful.</p>
<p><strong>“Everyone blames someone else.”</strong> <a href="https://xenospectrum.com/en/danfoss-freezer-cyberattack-second-wave/">The manufacturer</a> points to its customers&#39; passwords, while <a href="https://www.publicsenat.fr/actualites/politique/cyberattaque-contre-des-frigos-connectes-de-supermarches-des-systemes-de-refrigeration-exposes-sur-internet-alertent-des-experts-en-cybersecurite">the retail federation</a> calls for transparency from the manufacturer and the refrigeration engineers. Everything is true locally, but nothing describes the system as a whole.</p>
<h3 id="the-shortcuts">The shortcuts</h3>
<p><strong>A published vulnerability is not necessarily the door that was used.</strong> <a href="https://refindustry.com/fr/news/danfoss-copeland-et-trane-corrigent-des-failles-de-securite-des-controleurs/">Researchers had shown</a> that one Danfoss model generates a password of the day that can be deduced from its date, and then <a href="https://www.sudradio.fr/sud-radio/cyberattaque-contre-les-supermarches-cetait-dune-facilite-deconcertante">an expert invited on the radio</a> and <a href="https://www.journaldunet.com/cybersecurite/1556679-cyberattaques-sur-les-frigos-l-arbuste-qui-cache-la-foret/">the specialist press</a> saw in it the key to the attack. ANSSI, however, indicated that the attack had not exploited this vulnerability. The nuance matters, because a vulnerability is fixed with a patch, whereas a misconfiguration is fixed with an organisation.</p>
<p><strong>Three episodes do not necessarily make a campaign.</strong> France attributes October to Iran, pro-Russian groups claim November without evidence, and nothing links the EFS to any perpetrator. An open door attracts more than one visitor.</p>
<p><strong>Industrial IT is not business IT.</strong> There was talk of ransomware, of backups and of notifying the data protection authority, yet the attack neither stole nor encrypted any known data. <strong>The system was not blocked. It obeyed.</strong></p>
<h2 id="coming-next">Coming next</h2>
<p>In the hacked fridges affair, the main facts are known, and so are the remedies: do not expose, change the passwords, update. Why are they not applied? Not merely out of negligence, and these weaknesses are not specific to refrigeration either. <strong>The next article, <a href="/blog/why-the-fridges-stayed-open">“Why the fridges stayed open”</a>, lists the seven weaknesses that added up, and explains what makes each of them hard to close.</strong></p>
<p><em>Do you operate or maintain remotely maintained installations and see other constraints? Write to us at <a href="mailto:contact@trust-buster.com">contact@trust-buster.com</a>.</em></p>
<h2 id="sources">Sources</h2>
<p>Authorities:</p>
<ul>
<li><a href="https://cyber.gouv.fr/actualites/campagnes-de-d%C3%A9stabilisation-et-sabotage-men%C3%A9es-%C3%A0-lencontre-des-int%C3%A9r%C3%AAts-fran%C3%A7ais-par-la-r%C3%A9publique-islamique-diran/">ANSSI, destabilisation and sabotage campaigns carried out by the Islamic Republic of Iran (9 November 2026)</a>.</li>
<li><a href="https://cert.cyber.gouv.fr/alerte/CERTFR-2026-ALE-012">CERT-FR, alert CERTFR-2026-ALE-012</a>.</li>
<li><a href="https://cert.cyber.gouv.fr/uploads/CERTFR-2026-CTI-009.pdf">CERT-FR, report of the Cyber Crisis Coordination Centre CERTFR-2026-CTI-009 (9 November 2026)</a>.</li>
<li><a href="https://us.diplomatie.gouv.fr/en/declaration-attribution-iran-cyber-attacks-against-france">French Ministry for Europe and Foreign Affairs, attribution statement</a>.</li>
</ul>
<p>French press:</p>
<ul>
<li><a href="https://www.publicsenat.fr/actualites/politique/des-defauts-de-securisation-elementaire-et-des-mots-de-passe-faibles-ont-permis-la-cyberattaque-contre-les-frigos-de-supermarches">Public Sénat, hearing of the Director General of ANSSI (21 October 2026)</a>.</li>
<li><a href="https://www.techniques-ingenieur.fr/actualite/breves-afp/la-cyberattaque-des-frigos-de-supermarches-en-france-liee-a-des-defauts-de-securisation-elementaire-selon-lanssi-165159/">AFP via Techniques de l&#39;Ingénieur, basic security failings</a>.</li>
<li><a href="https://selectra.info/telecom/actualites/marche/efs-cyberattaque-chambres-froides-sang">Selectra, the EFS cold rooms</a>.</li>
<li><a href="https://www.lequotidiendumedecin.fr/sante-societe/des-chambres-froides-de-letablissement-francais-du-sang-victimes-dune-cyberattaque-dun-prestataire">The EFS, quoted by Le Quotidien du Médecin</a>.</li>
<li><a href="https://fr.euronews.com/2026/11/09/la-france-accuse-liran-de-cyberattaques-contre-les-frigos-de-supermarches">Euronews, France accuses Iran</a>.</li>
<li><a href="https://next.ink/266516/cyberattaque-contre-les-frigos-de-700-supermarches-pourquoi-la-france-accuse-liran/">Next, why France accuses Iran</a>.</li>
<li><a href="https://selectra.info/telecom/actualites/marche/frigos-supermarches-pirates-vague-europe">Selectra, the second wave and a Belgian installer</a>.</li>
<li><a href="https://www.24matins.fr/frigos-pirates-en-supermarche-lanssi-denonce-des-erreurs-basiques-1422863">24matins, the 2014 ANSSI guide</a>.</li>
<li><a href="https://www.publicsenat.fr/actualites/politique/cyberattaque-contre-des-frigos-connectes-de-supermarches-des-systemes-de-refrigeration-exposes-sur-internet-alertent-des-experts-en-cybersecurite">Public Sénat, experts and the retail federation</a>.</li>
<li><a href="https://refindustry.com/fr/news/danfoss-copeland-et-trane-corrigent-des-failles-de-securite-des-controleurs/">Refindustry, Danfoss and Copeland vulnerabilities fixed</a>.</li>
<li><a href="https://www.sudradio.fr/sud-radio/cyberattaque-contre-les-supermarches-cetait-dune-facilite-deconcertante">Sud Radio, “it was disconcertingly easy”</a>.</li>
<li><a href="https://www.journaldunet.com/cybersecurite/1556679-cyberattaques-sur-les-frigos-l-arbuste-qui-cache-la-foret/">Journal du Net, the shrub that hides the forest</a>.</li>
<li><a href="https://www.franceinfo.fr/internet/securite-sur-internet/cyberattaques/cyberattaque-contre-des-frigos-de-supermarches-le-parquet-de-paris-ouvre-une-enquete-visant-une-unite-des-gardiens-de-la-revolution-iraniens_8274806.html">Franceinfo, judicial investigation by the Paris prosecutor&#39;s office</a>.</li>
</ul>
<p>International press and analyses:</p>
<ul>
<li><a href="https://hard2bit.com/en/blog/french-supermarket-cyberattack-refrigeration-danfoss/">Hard2bit, affected Danfoss models and end of life</a>.</li>
<li><a href="https://www.helpnetsecurity.com/2026/11/02/european-supermarkets-refrigeration-units-cyberattacks/">Help Net Security, the second wave in Europe</a>.</li>
<li><a href="https://xenospectrum.com/en/danfoss-freezer-cyberattack-second-wave/">Xenospectrum, Danfoss statement on the Netherlands</a>.</li>
<li><a href="https://cybernews.com/security/russian-hackers-hacked-european-refrigeration-systems/">Cybernews, claim by NoName057(16) and Z-Pentest</a>.</li>
<li><a href="https://www.analyticsinsight.net/news/us-agencies-warn-of-cyberattacks-on-industrial-refrigeration-systems">Analytics Insight, exposure according to Censys</a>.</li>
<li><a href="https://www.bitsight.com/blog/supermarket-refrigeration-cyberattacks-data">Bitsight, data on attacks against supermarket refrigeration supervisors</a>.</li>
<li><a href="https://english.news.cn/20261111/9c874659cb83487db27ad65234968067/c.html">Xinhua, Iran rejects the accusations (10 November 2026)</a>.</li>
</ul>
<p><em>This article provides general information and our analysis, not advice tailored to your installation. The Cube and its satellites, presented later in this series, are the solution we recommend. Any change to a refrigeration installation is decided with your refrigeration engineer, after checking the manufacturer&#39;s warranty, the maintenance contract and your insurance.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Trackers, Bugs, Cameras: What Radio Listening Reveals</title>
      <link>https://trust-buster.com/blog/trackers-bugs-and-cameras-what-radio-listening-reveals</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/trackers-bugs-and-cameras-what-radio-listening-reveals</guid>
      <pubDate>Wed, 04 Nov 2026 18:00:00 GMT</pubDate>
      
      <description>These objects nearly all betray themselves by their emissions, and here is what permanent radio listening detects, what escapes it and what the law allows.</description>
      <content:encoded><![CDATA[<p>Twenty years ago, planting a bug in a meeting room required specialised equipment and some know-how, while today a Bluetooth tracker is sold in supermarkets, a bug that calls back a number as soon as it hears a voice sells online for the price of a meal, and a Wi-Fi camera fits inside a phone charger. Most of these objects were designed for perfectly legitimate uses, finding one&#39;s keys or watching one&#39;s home, which is precisely what makes them so easy to misuse, but nearly all of them betray themselves through their emissions, and that is what permanent radio listening can pick up.</p>
<p><em>This article is for managers of sensitive sites, meeting rooms and technical rooms wondering what radio monitoring can do against such objects.</em></p>
<h2 id="three-families-of-objects-three-signatures-from-the-bluetoot">Three families of objects, three signatures, from the Bluetooth tracker to the Wi-Fi camera</h2>
<p>All these objects share a feature that is also their weakness, because to be useful to whoever planted them they must transmit what they capture, and any wireless transmission leaves a trace in the radio spectrum, that is in the whole range of frequencies on which one can transmit, provided one knows where to look.</p>
<p><strong>Location trackers.</strong> Consumer Bluetooth trackers, such as Apple&#39;s AirTag, have no connection of their own, because they emit a short Bluetooth Low Energy, or BLE, advertisement which passers-by&#39;s phones relay to the manufacturer&#39;s network, so they are both discreet and talkative, since they announce their presence at short and regular intervals. Trackers with their own cellular modem, more expensive, transmit far less often, in brief bursts on mobile bands.</p>
<p><strong>Bugs.</strong> The most common use a SIM card and the mobile network, so they stay silent and then call or transmit as soon as a sound is detected, while others transmit continuously on free radio frequencies, like a small FM transmitter, and some record to a memory card to transmit only later, or never.</p>
<p><strong>Cameras.</strong> Most now use Wi-Fi, either joining an existing network or creating their own, and analogue models, still common in cheap equipment and in the drone world, transmit a picture continuously on frequencies such as 1.2 or 5.8 GHz.</p>
<h2 id="what-permanent-radio-listening-can-do">What permanent radio listening can do</h2>
<p>Radio listening installed permanently, like that of the Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), has over a one-off sweep, that is a search done once with a portable receiver, an advantage we consider decisive, because it knows the usual landscape of the place, it knows which devices are present every day, at what times and at what signal strength, and it is the deviation from this baseline that interests it. It receives and does nothing but receive, since it never tries to connect to what it hears nor to disturb it.</p>
<p><strong>Spotting a new Bluetooth transmitter that stays.</strong> A tracker forgotten in a bag crosses the building with its owner, while a tracker placed under a desk or attached to a company vehicle stays there, day after day, and the persistence of an unknown transmitter, matching no enrolled token, is a more reliable signal than its mere appearance. Apple and Google have also submitted to the IETF a <a href="https://datatracker.ietf.org/doc/draft-detecting-unwanted-location-trackers/">joint text</a> for detecting unwanted trackers, which makes some of these objects easier to recognise. That initial text has expired. The work continues in the <a href="https://datatracker.ietf.org/wg/dult/documents/">DULT working group</a>, whose accessory protocol reached its second version on 27 October 2026 without yet being published as an RFC.</p>
<p><strong>Spotting a Wi-Fi network or device that has no business there.</strong> An access point appearing in a meeting room, or a device continuously sending a lot of data from a fixed spot, deserves a check, and we described in <a href="/blog/a-wi-fi-honeypot-that-grants-nothing">A Wi-Fi Honeypot That Grants Nothing</a> what the Cube does with a probe or a fake network aimed directly at it.</p>
<p><strong>Spotting an emission where there should be none.</strong> On other bands, wider listening, entrusted to a software-defined radio, or SDR, can establish that an emission appears at an unusual frequency and time, even without being able to tell what it carries, and without trying to find out.</p>
<p>We went further into how the Cube listens to these bands, and how it detects attempts to blind or deceive it, in <a href="/blog/who-is-in-the-room">Who Is in the Room?</a>, <a href="/blog/what-an-sdr-finds">What an SDR Finds</a> and <a href="/blog/jamming-relays-and-the-flipper-zero">Jamming, Relays and the Flipper Zero</a>.</p>
<h2 id="what-escapes-it">What escapes it</h2>
<p>We think it essential to be as precise about limits as about capabilities, because a false sense of security seems to us worse than no protection.</p>
<ul>
<li><strong>An object that does not transmit cannot be seen.</strong> A bug recording to a memory card, a switched-off device or a tracker in deep sleep leave no radio trace, and finding them requires other techniques, such as non-linear junction detection or physical inspection, which are the trade of countermeasure specialists, or TSCM for technical surveillance countermeasures, whose job is precisely to look for these objects on site.</li>
<li><strong>A very short burst can slip between two measurements.</strong> A cellular bug transmitting for only a few seconds, while the listener looks elsewhere, can escape a sweep, and permanent listening reduces this risk without removing it.</li>
<li><strong>An emission is not a confession.</strong> A building&#39;s spectrum is saturated with legitimate emissions, and only persistence, location and context distinguish a suspicious object from a visitor&#39;s phone.</li>
<li><strong>Locating requires legwork.</strong> A fixed listener says something is happening within its perimeter, but finding the object then means moving around with a portable receiver and a directional antenna.</li>
</ul>
<h2 id="what-the-law-allows-and-what-it-forbids">What the law allows, and what it forbids</h2>
<p>Detecting transmitters in one&#39;s own premises does not seem to us to raise any particular legal difficulty, and it is even a precaution that the protection of sensitive information justifies, but two limits must be respected, and we give the French references here, since similar rules exist in other European countries.</p>
<p>First, <strong>detecting is not listening</strong>, because capturing, recording or transmitting words spoken in private without their author&#39;s consent is an offence (Article 226-1 of the French Criminal Code), as is intercepting correspondence (Article 226-15), so legitimate radio monitoring is interested in the existence of emissions, their frequency and strength, never in their content.</p>
<p>Second, the manufacture, import, possession and sale of devices designed to carry out such interceptions are regulated and subject to authorisation (Article 226-3 of the French Criminal Code), and the receivers we use are general-purpose radio equipment, which measure and do not demodulate third parties&#39; communications, and we intend to keep it that way.</p>
<p>Finally, like any device installed in premises where employees work, permanent radio listening must be presented to the works council and brought to people&#39;s attention, even if it does not target their activity, and we discussed this in <a href="/blog/detecting-presence-without-monitoring-employees">Detecting Presence Without Monitoring Employees</a>.</p>
<h2 id="where-to-start">Where to start</h2>
<p>If the question arises for your site, we would suggest starting by defining the places that really matter, a boardroom, a laboratory or a technical room, installing radio listening there for a few weeks in plain observation to establish the baseline, and only then deciding which alerts to arm. For an in-depth search, notably before a sensitive meeting, calling on a countermeasure specialist remains the right answer, and permanent listening will be valuable to them, since it will tell them what has changed since their last visit.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>IETF, <em>Detecting Unwanted Location Trackers</em>, initial draft submitted by Apple and Google (draft-detecting-unwanted-location-trackers, last version 20 December 2023, expired). <a href="https://datatracker.ietf.org/doc/draft-detecting-unwanted-location-trackers/">datatracker.ietf.org</a></li>
<li>IETF, DULT working group, <em>Detecting Unwanted Location Trackers</em>. <a href="https://datatracker.ietf.org/wg/dult/documents/">datatracker.ietf.org</a></li>
<li>French Criminal Code, Articles 226-1, 226-3 and 226-15, and Articles R226-1 et seq. (list and authorisation of devices).</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>USB and Serial Links: The Door Left Open</title>
      <link>https://trust-buster.com/blog/usb-and-serial-links-the-door-left-open</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/usb-and-serial-links-the-door-left-open</guid>
      <pubDate>Sun, 01 Nov 2026 18:00:00 GMT</pubDate>
      
      <description>Off the network, USB and serial links are often the only way in and the least watched, and here is what can be done without touching the machine.</description>
      <content:encoded><![CDATA[<p>It is commonly said that equipment not connected to the network is protected from cyberattacks, and that is partly true, since it is a decision many operators take knowingly, but it forgets that this equipment almost always has another door, more discreet and often open to all, which lives in the USB and serial links of the machine, because there is a USB port on the front of the operator panel, a serial or Ethernet programming port on the controller, and a maintenance connector in the cabinet. Updates, backups and settings go through it, and so can everything else.</p>
<p><em>This article is for industrial systems security managers and maintenance teams running a fleet of equipment isolated from the network.</em></p>
<h2 id="an-interface-that-trusts-whatever-is-plugged-in">An interface that trusts whatever is plugged in</h2>
<p>The USB port was designed for convenience, because a device receiving a peripheral asks it what it is and takes its word for it, whether it presents itself as a storage stick, a keyboard, a network card or a serial adapter, and nothing in the protocol lets it check that the object is what it claims to be.</p>
<p>In 2014, researchers showed that the microcontroller of an ordinary USB stick could be reprogrammed so that, in addition to storage, it presents itself as a keyboard able to type commands far faster than a human hand, and this family of attacks, known since as BadUSB, has given rise to freely sold tools, keystroke-injection sticks, charging cables that hide a radio transmitter, innocent-looking adapters, which require no particular skill and which, on equipment without a log, leave no trace.</p>
<p>Four years earlier, the Stuxnet worm had already shown, on another scale, that an installation isolated from the network, what is also called an air gap, could be reached through removable media and the engineering workstations that read them, so network isolation greatly reduces the attack surface but it does not remove it, because it shifts it to whatever gets plugged in.</p>
<h2 id="the-serial-port-older-and-even-more-permissive">The serial port, older and even more permissive</h2>
<p>On many controllers, the small industrial computers that drive the machine, and on much older equipment, the programming port asks for no password, or a default password known to the whole trade, and whoever connects to it with the right software can read and modify the program, change control parameters or force outputs. This is not a design flaw by the standards of the time, because the port was in the cabinet, the cabinet was locked, and the person opening it was the automation engineer.</p>
<p>It is exactly that assumption that no longer holds, or not always, because the cabinet opens with a key bought online, the automation engineer is often a contractor, and the laptop they plug in spent the previous day on other sites.</p>
<h2 id="what-the-frameworks-recommend-from-anssi-to-iec-62443">What the frameworks recommend, from ANSSI to IEC 62443</h2>
<p>Recommendations converge, and they are not new.</p>
<p><strong>ANSSI</strong>, the French national cybersecurity agency, in its detailed measures for industrial control system security, recommends controlling the use of removable media, disabling unused ports, dedicating maintenance workstations to industrial systems and passing media through a decontamination station before any use.</p>
<p><strong>IEC 62443-3-3</strong> devotes a specific requirement to use control for portable and mobile devices, referenced SR 2.3, and another to malicious code protection, referenced SR 3.2, notably asking to restrict the use of removable media and to log their use.</p>
<p>These recommendations are sound, and we endorse them, but their difficulty lies in applying them to the existing base, because disabling a port often means changing the equipment&#39;s configuration, which the manufacturer does not always allow, and a decontamination station only protects against media people have the discipline to pass through it.</p>
<h2 id="interposing-on-usb-and-serial-links">Interposing on USB and serial links</h2>
<p>For interfaces as for power, we preferred an approach that does not touch the equipment and that consists in interposing on the link, at its boundary, because a box placed between the port and whatever is plugged into it can, without modifying the machine, apply a few simple rules.</p>
<ul>
<li><strong>Only allow the link on the decision of the Cube</strong>, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), that is for the right person, on the right equipment, during the planned slot, as for powering up.</li>
<li><strong>Restrict what can present itself</strong> on a USB port, for instance refusing a keyboard where only a storage stick is expected.</li>
<li><strong>Log every connection</strong>, including refused ones, with the identity of the person present at that moment.</li>
</ul>
<p>Such a box does not read the content of files and therefore does not replace the decontamination station, because it only decides who plugs in, when, and in what form, and that is already what is most often missing. We must also be honest about the state of our work, because the effector that is fully operational today, that is the box which applies the Cube&#39;s decision, is the one acting on the power supply, while the satellites for USB and serial links are under development and not yet offered as products, and we would rather say so clearly than suggest otherwise.</p>
<h2 id="what-can-be-done-right-now">What can be done right now</h2>
<p>Meanwhile, and whatever device is chosen, a few simple measures reduce the risk on an offline fleet.</p>
<ol>
<li><strong>Inventory the physical interfaces</strong> of each critical piece of equipment, USB and serial links included, noting which ports exist, who can reach them and what they are used for.</li>
<li><strong>Physically close what is not used</strong>, with lockable port blockers and cabinets locked with a lock whose keys are tracked rather than a universal triangle key.</li>
<li><strong>Dedicate maintenance equipment</strong>, that is a laptop and media reserved for industrial systems, used for nothing else.</li>
<li><strong>Pass media through a decontamination station</strong> before any connection, and keep a record.</li>
<li><strong>Track who opens the cabinet</strong>, which is often a simple and yet neglected measure, and that is where access control at the equipment boundary joins the topic.</li>
</ol>
<p>For the particular case of the contractor who comes to plug a laptop into the controller&#39;s port, see also <a href="/blog/external-contractors-who-touches-your-machines">External Contractors: Who Touches Your Machines?</a>.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>Nohl, K. and Lell, J. <em>BadUSB: On accessories that turn evil</em>. Black Hat USA, 2014. <a href="https://infocondb.org/con/black-hat/black-hat-usa-2014/badusb-on-accessories-that-turn-evil">infocondb.org</a></li>
<li>Falliere, N., Murchu, L. O. and Chien, E. <em>W32.Stuxnet Dossier</em>. Symantec, version 1.4, February 2011.</li>
<li>ANSSI, <em>La cybersécurité des systèmes industriels, mesures détaillées</em>, 2014. <a href="https://cyber.gouv.fr/publications/la-cybersecurite-des-systemes-industriels">cyber.gouv.fr</a></li>
<li>IEC 62443-3-3:2013, System security requirements and security levels (SR 2.3, SR 3.2).</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>Detecting Presence Without Monitoring Employees</title>
      <link>https://trust-buster.com/blog/detecting-presence-without-monitoring-employees</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/detecting-presence-without-monitoring-employees</guid>
      <pubDate>Wed, 28 Oct 2026 18:00:00 GMT</pubDate>
      
      <description>Detecting presence near a machine touches on personal data and on the monitoring of employees, and we set out what to settle before installing anything.</description>
      <content:encoded><![CDATA[<p>Machine access control rests, by construction, on sensitive information, since it works by detecting presence, that of an identified person at a given place and time, and that is what makes it useful, but it is also what brings it close, if one is not careful, to an employee monitoring tool, because the technology is the same in both cases and the difference lies in the purpose, the rules of use and a few design choices best made before installation rather than after the first works council meeting.</p>
<p><em>This article is for operators, HR directors, data protection officers and staff representatives reviewing a machine access control project.</em></p>
<h2 id="what-the-device-actually-collects">What the device actually collects</h2>
<p>It is worth starting with a precise description of what is recorded, because concerns often stem from a broader idea than reality. The work of a Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), consists in detecting presence, since it establishes that an enrolled token, badge or phone, is near a machine, verifies its identity through a cryptographic exchange, and records the decisions it takes, grant, denial, alert, bypass attempt, but it does not film, records no sound, does not locate anyone outside its perimeter and uses no biometric data.</p>
<p>What it produces is nonetheless personal data as soon as it is tied to a named person, and when presence is crossed with machine use to infer what someone did and for how long, one enters the field of monitoring employees&#39; activity, which has its own rules.</p>
<h2 id="what-the-law-requires">What the law requires</h2>
<p>Taking France as an example, three bodies of rules overlap, as we understand them without being lawyers, and none of the three seems to us something one can leave aside, while other EU countries have, as far as we know, comparable rules, often stricter regarding works councils.</p>
<p><strong>Labour law.</strong> No information concerning an employee personally may be collected by a device not previously brought to their attention (Article L1222-4 of the French Labour Code), restrictions on employees&#39; freedoms must be justified by the nature of the task and proportionate to the aim pursued (Article L1121-1), and the works council (CSE) must be informed and consulted, before they are implemented, on the means and techniques allowing employees&#39; activity to be monitored (Article L2312-38). The word before matters, because on our reading a device installed prior to that consultation is hard to regularise afterwards, and what it collected may not be usable against employees.</p>
<p><strong>The GDPR.</strong> Processing must pursue a specified and legitimate purpose, collect only necessary data, keep it for a limited period and inform the people concerned, and for badge access control to premises the French data protection authority (CNIL) states that access data should be deleted three months after it is recorded, which gives a useful order of magnitude, even though a specific evidential purpose may justify a different period, provided it is documented.</p>
<p><strong>The impact assessment.</strong> The CNIL lists among processing operations requiring a data protection impact assessment, the DPIA, those whose purpose is to constantly monitor employees&#39; activity, and a well-designed access control does not have that purpose, but it is prudent to demonstrate it in writing rather than assume it, and the impact assessment is often the best document for doing so.</p>
<h2 id="a-question-of-purpose">A question of purpose</h2>
<p>Everything hinges on purpose, and that is what must be stated clearly before the works council, because some purposes are in our view legitimate and proportionate, such as ensuring a dangerous machine is only used by trained people, detecting use outside a slot, keeping track of the <a href="/blog/external-contractors-who-touches-your-machines">external contractors</a> who come to your machines or billing the use of shared equipment, while others are not, and we refuse them on principle, such as ranking individual productivity, measuring break times or tracking a person&#39;s movements through the building.</p>
<p>A charter forbidding these uses seems to us less protective than a device that does not make them easy to carry out, and that is therefore what we sought to achieve in design rather than in usage policy.</p>
<h2 id="detecting-presence-without-reconstructing-activity">Detecting presence without reconstructing activity</h2>
<p><strong>Aggregate by default.</strong> Reports are produced per machine, per period and per organisation, and named attribution only appears on a trigger defined in advance, use outside a slot, anomaly, bypass attempt or justified audit request.</p>
<p><strong>Distinct retention periods.</strong> Named data has a short retention period, set with you, whereas machine consumption data, which no longer has anything personal once detached from presence, follows your management needs.</p>
<p><strong>No tracking outside the perimeter.</strong> A Cube knows nothing of what happens beyond its range and does not try to reconstruct a route, because it answers a narrow question, which is whether this person, authorised or not, is here, in front of this machine.</p>
<p><strong>Data stays on site by default.</strong> Decisions are taken and data is processed locally, in the Cube, and nothing leaves for a server unless you have decided so, which greatly simplifies the record of processing and the question of transfers.</p>
<p><strong>The device announces itself.</strong> Signage near protected equipment and individual information at enrolment are part of the deployment, just like installing the sockets.</p>
<h2 id="an-order-for-doing-things">An order for doing things</h2>
<p>If you are considering such a device, we would suggest proceeding in this order, which avoids most of the difficulties we have seen in the field.</p>
<ol>
<li><strong>Write down the purpose</strong> in one sentence, with the list of excluded uses.</li>
<li><strong>Involve the data protection officer</strong>, the DPO, from the start, and carry out the impact assessment if it is required or if there is any doubt.</li>
<li><strong>Inform and consult the works council</strong> before any installation, describing what is collected, what is not and the retention periods.</li>
<li><strong>Start in Observation mode</strong>, which cuts nothing and lets you show the works council, with data, what detecting presence means in practice and what the device actually sees.</li>
<li><strong>Inform each enrolled person individually</strong>, employee or external contractor.</li>
</ol>
<p>A project presented this way has been well received in the cases we have encountered, because it answers a concern teams share, which is knowing that the dangerous machine next door cannot be started by just anyone, and it is an occupational safety argument before it is a control argument, so we believe it should be presented as such.</p>
<p>For what the Cube knows, and will never know, about the use of the machines themselves, see <a href="/blog/what-the-cube-knows-about-your-machines">What the Cube Knows About Your Machines</a>, and for radio monitoring specifically, see <a href="/blog/is-radio-monitoring-in-a-workplace-legal">Is Radio Monitoring in a Workplace Legal?</a>.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>French Labour Code, Articles L1121-1, L1222-4 and L2312-38.</li>
<li>Regulation (EU) 2016/679 (GDPR), Articles 5, 6, 13 and 35. <a href="https://eur-lex.europa.eu/eli/reg/2016/679/oj">EUR-Lex</a></li>
<li>CNIL, <em>Exemple d&#39;information en cas d&#39;accès à des locaux professionnels par badge</em>. <a href="https://cnil.fr/fr/RGPD-exemple-information-acces-aux-locaux-professionnels-par-badge">cnil.fr</a></li>
<li>CNIL, <em>Liste des types d&#39;opérations de traitement pour lesquelles une analyse d&#39;impact est requise</em> (deliberation No 2018-327 of 11 October 2018). <a href="https://www.cnil.fr/sites/default/files/atoms/files/liste-traitements-avec-aipd-requise-v2.pdf">cnil.fr</a></li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>Shared Workshops: Who Answers for the Machine?</title>
      <link>https://trust-buster.com/blog/shared-workshops-who-answers-for-the-machine</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/shared-workshops-who-answers-for-the-machine</guid>
      <pubDate>Sun, 25 Oct 2026 18:00:00 GMT</pubDate>
      
      <description>In shared workshops, fab labs, schools or laboratories, the question of liability comes before security, and that question gave birth to the Cube.</description>
      <content:encoded><![CDATA[<p>The Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), was not born in a factory, nor in a defence programme, but in one of those shared workshops where dangerous machines are available to dozens of people, and more precisely in a question we asked ourselves when we became responsible for one of them, because someone has to know who answers for what happens there, and how anyone knows.</p>
<p><em>This article is for managers of fab labs, makerspaces, school and university workshops, laboratories and shared technical platforms, for the chairs of the associations running them, and for their insurers.</em></p>
<h2 id="sofab-the-sophia-antipolis-fab-lab">SoFAB, the Sophia Antipolis fab lab</h2>
<p>In October 2014, the Telecom Valley association opened SoFAB, the Sophia Antipolis fab lab, in the premises of Polytech Nice Sophia on the SophiaTech campus, together with the University of Nice Sophia Antipolis, the Pobot robotics association and a dozen partners from the local ecosystem. The project had matured for two years and it was led by Pascal Flamand, then vice-president of Telecom Valley in charge of the fab lab, and the place brought together 3D printers, a CNC milling machine, a laser cutter, a vinyl cutter and professional workbenches, and it was deliberately open to all, students, secondary school pupils, project leaders, employees of small firms and large groups, the general public, with time slots for each audience.</p>
<p>On 31 March 2015, Pascal Flamand became president of Telecom Valley, and with the role came a responsibility that is hard to gauge until one takes it on, that of the association making machine tools available to an open public, and that of its president, because the laser cutter can catch fire, the milling machine can injure, and a teenager who came to discover 3D printing can walk past a machine they never learned to use. As in any fab lab, rules, training and supervision by the fab manager and volunteers exist, but at the moment a machine starts, nothing tells whether the person using it has been trained, whether it is their slot, or even who they are.</p>
<p>An answer was needed that relied neither on locking the machines away, nor on a supervisor&#39;s permanent presence, nor on trust alone, and that search gave birth to the Cube, devised by Julien Holtzer, a volunteer at the fab lab, together with Pascal Flamand, and it is the same question, transposed to industry, that we address today.</p>
<h2 id="a-place-where-liability-dissolves">A place where liability dissolves</h2>
<p>A shared workshop brings together conditions rarely found together in industry, because users are many and change often, their skills range from yesterday&#39;s beginner to the seasoned professional, opening hours are wide, sometimes without supervision, and the machines were not designed to know who is using them, and to this one readily adds a culture of trust and mutual help, which is the charm of these places and makes setting rules all the more delicate.</p>
<p>In this context liability does not disappear but it dissolves, because it is shared between the association running the place, its president, the institution hosting it where applicable, the supervisors, the trainers and each user, and the day an accident happens it must nonetheless be reconstructed, and that is when one discovers what one should have been able to show.</p>
<h2 id="what-other-shared-workshops-have-been-through">What other shared workshops have been through</h2>
<p>Serious accidents remain rare in shared workshops, but those that are documented look strikingly alike, and they are all about access as much as machines.</p>
<p><strong>Vienna, spring 2011.</strong> At Happylab, one of Europe&#39;s first fab labs, a member starts a job on a new laser cutter and goes to lunch, then the machine&#39;s axis gets stuck, the laser heats the same spot for a long time, the acrylic catches fire, then the lid and the ceiling, and six fire engines are called, the machine is destroyed and much of the space is burned or covered in soot, but nobody is hurt. Rather than restricting opening hours, the founders design a box that controls the machines&#39; power and requires the user to authenticate regularly, or the machine stops, and that device later became a product in its own right, so we are not the first to conclude that the power supply is the right place to enforce a rule of use, and we find it rather reassuring that others got there by another route.</p>
<p><strong>New Haven, April 2011.</strong> A Yale University student dies after her hair is caught in a lathe in a laboratory&#39;s machine shop, where she was working alone late in the evening, and in its findings the US Occupational Safety and Health Administration notes that the lathe lacked an emergency stop button able to cut power, that the rules for using the equipment were not posted, and recommends that students do not work alone, after which the university immediately launches a review of the rules of all its workshops with power equipment.</p>
<p><strong>In industry too.</strong> The ARIA database, which records technological accidents in France, mentions for instance a fire that broke out in 2017 on a laser cutting machine in a plastics plant in the Meuse, where the fire spread to the 800 m² building, two employees were hospitalised for smoke inhalation and production stopped for two days.</p>
<p>These three cases differ in severity, but they raise the same question as ours, because one has to know who was in front of the machine, whether they were trained, whether they were alone, and whether the machine could run without anyone watching.</p>
<h2 id="liability-including-the-president-s">Liability, including the president&#39;s</h2>
<p>We are not lawyers and this article is no substitute for advice, but a few general principles of French law, as we understand them, shed light on the question, and most European countries follow, as far as we know, comparable rules.</p>
<p><strong>Custody of the thing.</strong> In civil matters, one is liable for damage caused by things in one&#39;s custody (Article 1242 of the French Civil Code), and in a shared workshop the question of who had custody of the machine at the time of the accident, the association or the user, is often the first one debated, and it depends largely on what governed its use.</p>
<p><strong>The duty of safety.</strong> Whoever provides a service must do so in conditions offering the safety one may legitimately expect (Article L421-3 of the French Consumer Code), and when the workshop employs staff, such as a fab manager, the French Labour Code also places a safety obligation on the employer (Article L4121-1), which involves assessing risks (Article L4121-3), providing suitable work equipment (Article R4321-1) and training those who use it.</p>
<p><strong>The president, the association&#39;s agent.</strong> The president of an association acts on its behalf, and in civil matters it is in principle the association, as a legal person, that compensates damage caused in the course of its activities, whereas the president only answers personally to the association for management faults (Article 1992 of the Civil Code, applied less strictly to an unpaid mandate), and towards third parties only for a fault separable from their duties, meaning one of particular gravity and incompatible with their normal exercise, of which failing to take out mandatory insurance is the classic example.</p>
<p><strong>The president before a criminal court.</strong> This is where the question seems to us most serious, because the criminal liability of the association does not exclude that of individuals (Article 121-2 of the French Criminal Code), and in case of injury or death a president who did not directly cause the harm but created the situation that allowed it, or failed to take the measures to prevent it, may be held liable if they committed a characterised fault, exposing others to a particularly serious risk they could not ignore (Article 121-3, as amended by the law of 10 July 2000). Leaving a dangerous machine accessible to all, without control and knowing the risk, seems to us to come close to that definition, even though it is for the court to decide in each case, and a delegation of powers to a fab manager can shift this liability, provided the delegate genuinely has the competence, authority and means required.</p>
<p>In all these cases, what makes the difference has less to do, in our view, with the existence of rules than with the ability to show that they were applied, because a signed charter and a training session weigh little if, in practice, anyone could start the machine, and house rules declaring the association &quot;not liable&quot; for accidents carry limited weight against a safety failure.</p>
<h2 id="the-usual-answers-and-their-limits">The usual answers, and their limits</h2>
<p>Shared workshops have not stood still, and one usually finds a stack of reasonable measures.</p>
<ul>
<li><strong>Mandatory training</strong>, sealed by a &quot;machine passport&quot; or internal qualification, is essential, but nothing stops an untrained person from pressing the button.</li>
<li><strong>Online booking</strong> of time slots organises use but does not control it, since the machine knows nothing of the schedule.</li>
<li><strong>Key or badge boxes</strong> on some machines work, but they are costly to generalise, often vendor-specific, and they usually do not know whether training is up to date.</li>
<li><strong>The presence of a supervisor</strong> remains the best protection, but no place can guarantee it at every hour.</li>
</ul>
<p>Each of these measures addresses part of the problem, but none links the three pieces of information that matter when the machine starts, because one has to know who is there, whether they have been trained on this machine, and whether it is their slot.</p>
<h2 id="what-the-cube-does-in-shared-workshops">What the Cube does in shared workshops</h2>
<p>That link is exactly what the Cube was designed to establish, and it is in this context that the logic of machine access control appears most simply, because each machine is powered through a controlled socket, and for it to receive power the Cube must establish the presence of an enrolled person, check that their training on that machine is valid, and that the use matches an authorised slot, so all three conditions are needed at once, failing which the socket stays off.</p>
<p>Two behaviours matter particularly in an open place, and the accidents above show why. When the authorised user walks away while the machine is still powered, an alert is raised and the machine can be protected against being restarted by someone else, and for a laser cutter left alone a rule can go as far as requiring the user&#39;s presence throughout the job, while when a rule requires two people for a machine, or forbids its use outside supervised hours, the Cube can check it rather than rely on a notice on the wall.</p>
<p>The log, finally, answers the question one always asks too late, who used this machine, when, and with what right, and in shared workshops it is in fact used far more often for billing usage and organising slots than for analysing incidents, which seems to us rather good news.</p>
<h2 id="what-it-does-not-change">What it does not change</h2>
<p>Access control does not make a machine less dangerous for the person entitled to use it, because it replaces neither training, nor the machine&#39;s guards, nor the emergency stop, nor the presence of a supervisor for uses that require one, and we detailed this distinction in <a href="/blog/power-off-is-not-an-emergency-stop">Cutting the Power Is Not an Emergency Stop</a>. Nor does it settle, on its own, the legal question of liability, but what it brings is the ability to show that the rules you had set were actually enforced by the machines themselves, and not just posted on the wall.</p>
<p>Members of shared workshops are also entitled, like everyone, to have data about them limited to what is necessary, kept for a reasonable time and brought to their attention, and we come back to this in <a href="/blog/detecting-presence-without-monitoring-employees">Detecting Presence Without Monitoring Employees</a>.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>WebTimeMedias, <em>SoFab à Sophia : un FabLab dédié à l&#39;Internet des Objets</em>, 17 October 2014. <a href="https://www.webtimemedias.com/article/sofab-sophia-un-fablab-dedie-linternet-des-objets">webtimemedias.com</a></li>
<li>Pôle SCS, <em>Lancement communautaire de SoFAB, le FabLab de Sophia Antipolis</em>, 21 November 2014. <a href="https://aktantis.com/article/lancement-communautaire-de-sofab-le-fablab-de-sophia-antipolis">aktantis.com</a></li>
<li>Invest in Côte d&#39;Azur, <em>SoFab: The Sophia Antipolis IoT FabLab</em>. <a href="https://www.investincotedazur.com/en/sofab-the-sophia-antipolis-fablab-dedicated-to-the-internet-of-things-iot/">investincotedazur.com</a></li>
<li>Le Journal des Entreprises, <em>Sophia Antipolis : Pascal Flamand prend la tête de Telecom Valley</em>, 2 April 2015. <a href="https://www.lejournaldesentreprises.com/article/sophia-antipolis-pascal-flamand-prend-la-tete-de-telecom-valley-106493">lejournaldesentreprises.com</a></li>
<li>Polytech Nice Sophia, <em>SoFab, le FabLab de Sophia-Antipolis</em>. <a href="https://polytech.univ-cotedazur.fr/fablab-sophia-antipolis">polytech.univ-cotedazur.fr</a></li>
<li>SoFAB by Telecom Valley. <a href="https://www.sofab.tv">sofab.tv</a></li>
<li>Fabman, <em>The Story of Fabman</em> (Happylab fire, Vienna, 2011). <a href="https://fabman.io/blog/2020-01-28-fabman-story">fabman.io</a></li>
<li>CBS News, <em>OSHA Faults Yale University In Fatal Lab Accident</em>, 16 August 2011. <a href="https://www.cbsnews.com/newyork/news/osha-faults-yale-university-in-fatal-lab-accident">cbsnews.com</a></li>
<li>BARPI, ARIA database, accident No 49555, <em>Feu sur une machine de découpe au laser</em>, 25 April 2017. <a href="https://www.aria.developpement-durable.gouv.fr/accident/49555/">aria.developpement-durable.gouv.fr</a></li>
<li>Associathèque, <em>Focus : la responsabilité du président</em>. <a href="https://www.associatheque.fr/fr/focus-responsabilite-president.html">associatheque.fr</a></li>
<li>Justice.fr, <em>Dommage causé par un dirigeant d&#39;association : qui est responsable ?</em> <a href="https://www.justice.fr/fiche/dommage-cause-dirigeant-association-responsable">justice.fr</a></li>
<li>French Civil Code, Articles 1242 and 1992.</li>
<li>French Criminal Code, Articles 121-2 and 121-3, and Law No 2000-647 of 10 July 2000.</li>
<li>French Consumer Code, Article L421-3.</li>
<li>French Labour Code, Articles L4121-1 and L4121-3 (safety obligation and risk assessment). <a href="https://www.legifrance.gouv.fr/codes/article_lc/LEGIARTI000035640828">Légifrance</a></li>
<li>French Labour Code, Article R4321-1 (work equipment).</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>A Proof Log: What It Proves, and What It Does Not</title>
      <link>https://trust-buster.com/blog/a-proof-log-what-it-proves-and-what-it-does-not</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/a-proof-log-what-it-proves-and-what-it-does-not</guid>
      <pubDate>Wed, 21 Oct 2026 17:00:00 GMT</pubDate>
      
      <description>A proof log that is timestamped, signed and chained can be produced before an auditor or a judge, provided you know exactly what it establishes.</description>
      <content:encoded><![CDATA[<p>We readily write that every grant, every denial and every bypass attempt produces an entry that is timestamped and justified, that each entry is chained to the previous one, and that the head of the chain is signed at every upload. The phrase is accurate, but it raises a question our most rigorous counterparts never fail to ask, because one needs to know signed by whom, timestamped by which clock, and above all what such a proof log actually establishes on the day it must be produced before someone who has no reason to take our word for it.</p>
<p><em>This article is for security managers, lawyers and auditors who must assess what an event log produced by a device is worth, ours or anyone else&#39;s.</em></p>
<h2 id="what-an-audit-trail-is-for-when-all-goes-well">What an audit trail is for, when all goes well</h2>
<p>The audit log, or audit trail, is often imagined as an exhibit for a trial, brought out the day things go wrong, whereas that is in fact its rarest use, since in the vast majority of cases it answers mundane questions, who used this bench last week, how many times was this machine started outside its slot this month, did the planned maintenance take place. It then serves, which is already less mundane, to show an auditor that a control exists and works, since an access policy that leaves no trace cannot be told apart, from the outside, from a policy that is not applied.</p>
<p>Only last does it become evidence in the strong sense, produced in a dispute with a contractor, before an insurer after a loss, or in proceedings, but it is this last use that dictates how it is designed, because a log that would not withstand adversarial examination is worth little more than a handover notebook.</p>
<h2 id="what-the-law-expects-of-an-electronic-record">What the law expects of an electronic record</h2>
<p>We are not lawyers and we report here what we understand of it. Under French law, the Civil Code gives an electronic record the same evidential value as a paper one, on two conditions, which are that the person it comes from can be duly identified and that it is created and kept in conditions that guarantee its integrity (Article 1366). Between merchants, and in criminal matters, evidence is free, meaning a judge may rely on any element whose weight they assess, provided it was obtained fairly and debated by both sides, and most European legal systems follow, as far as we know, comparable principles.</p>
<p>At European level, the eIDAS Regulation defines so-called qualified electronic signatures, seals and timestamps, issued by supervised trust service providers, which benefit from a presumption of integrity and accuracy, and we want to be clear on this, because the signature produced by a field device is not a qualified electronic seal under eIDAS and does not carry that legal presumption, so its value rests on what can be demonstrated about how it is produced, and that is exactly what interests us here.</p>
<h2 id="what-a-well-designed-proof-log-establishes">What a well-designed proof log establishes</h2>
<p>A proof log does not need to be perfect to be useful, but it must answer four questions precisely.</p>
<p><strong>Who produced this entry?</strong> Each Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), has its own cryptographic identity, meaning a key pair generated in the device and confined to a secure component, and at every retrieval the Cube signs the head of its log chain with this key, and since each entry is linked to the previous one, that signature covers every entry before it, so a valid signature establishes that the log comes from that Cube, and from no other device or intermediate server, provided its key has not been extracted. That is the form non-repudiation takes here.</p>
<p><strong>Has it been altered since?</strong> Each entry carries a fingerprint of its own content and of the entry before it, a fingerprint being a digest computed from the content whose value changes as soon as the content changes, which makes the log a hash chain, so changing a single character of an entry changes its fingerprint, breaks the chain from that point and invalidates the signature, and the result is a tamper-evident log that a third party can check with the Cube&#39;s public key, without having to take our word for it.</p>
<p><strong>Are entries missing?</strong> This is the most often forgotten question, and yet it is where conventional logs seem to us weakest, because one can rarely modify a line unnoticed, but one can often delete one. Chaining answers this in part, since a deletion in the middle of the log breaks the chain detectably, but a cut at the end of the log would not break it, and that is why every entry carries a sequential number and why, at each retrieval, the head of the chain is handed to the authority, which acknowledges it, because rewriting the past would then also mean rewriting what the server has already received. We preferred the log to be able to say by itself that something is missing at a given point, rather than leave the reader to guess.</p>
<p><strong>When?</strong> The timestamp comes from the Cube&#39;s clock, set by an authenticated holder during synchronisation, and every clock setting is itself a log entry, whereas an order arriving while the clock is not deemed trustworthy is refused rather than executed with a doubtful date, so accuracy depends on the last synchronisation, but the uncertainty is known rather than hidden.</p>
<h2 id="what-it-does-not-prove">What it does not prove</h2>
<p>It would be tempting to stop there, but we think, on the contrary, that the value of evidence lies as much in the list of what it establishes as in the honesty with which its limits are described, and that is, in our opinion, what an expert would look at first.</p>
<p><strong>It proves that a credential was presented, not that a person was there.</strong> The log establishes that an enrolled token, badge or phone, answered a cryptographic challenge correctly, and if the token was lent, stolen or used under duress, the log will not say so, which is true of any access control, including the building badge, and that is why we never claim to identify a person in the strict sense.</p>
<p><strong>It proves what the device saw, not what happened.</strong> A machine powered through an extension lead from another socket, outside the device&#39;s perimeter, will not appear in the log, or only indirectly as an anomaly, because the log describes the protected perimeter, and nothing beyond it.</p>
<p><strong>It depends on the integrity of the device producing it.</strong> A signature guarantees that an entry comes from the Cube, but it does not guarantee that the Cube was not deceived or compromised, and that is why the key is confined to a secure component rather than ordinary memory, why physical tampering attempts are themselves logged events, and why we publicly describe the attacks we know how to counter and those we cannot yet counter.</p>
<p><strong>Its timestamp is not a qualified timestamp.</strong> It is consistent, monotonic and documented, but it carries no legal presumption of accuracy, and when a date must be asserted against a third party with certainty, it remains possible to have the exported log timestamped by a qualified provider, for instance when it is uploaded to the server.</p>
<h2 id="what-makes-the-difference-before-a-third-party">What makes the difference before a third party</h2>
<p>In the cases we have encountered, what gives a proof log weight before an auditor or an expert has less to do with cryptography than with three very concrete qualities. The first is continuity, because a log covering the whole period, with no unexplained gaps, including during a network outage of several days, is worth more than a richer but patchy one. The second is symmetry, because a denial must be recorded as carefully as a grant, otherwise the log only tells what went well. The third is reproducible verification, because a third party must be able to check signatures and chaining with standard tools, without depending on the vendor, which requires formats and public keys to be documented.</p>
<p>These are the three qualities we prioritised, and we invite you to judge us on them rather than on the word &quot;signed&quot;, which on its own says very little.</p>
<h2 id="a-few-precautions-on-your-side">A few precautions on your side</h2>
<p>A proof log remains data you keep, and its value also depends on how you handle it, so we would suggest setting a retention period suited to your obligations and to the personal nature of the data it contains, keeping exported logs in a space whose access is itself logged, and never reworking the original files, since analysis is done on a copy and the original must remain verifiable. These precautions are more common sense than technology, but they are what make it possible, eighteen months later, to say without hesitation that the document produced is the one the device wrote, and we say what this log contains that is personal, and how to limit it, in <a href="/blog/detecting-presence-without-monitoring-employees">Detecting Presence Without Monitoring Employees</a>.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>French Civil Code, Articles 1366 and 1367 (electronic records and signatures).</li>
<li>French Commercial Code, Article L110-3, and Code of Criminal Procedure, Article 427 (freedom of evidence).</li>
<li>Regulation (EU) No 910/2014 (eIDAS), as amended by Regulation (EU) 2024/1183. <a href="https://eur-lex.europa.eu/eli/reg/2014/910/oj">EUR-Lex</a></li>
<li>ISO/IEC 27037:2012, Guidelines for identification, collection, acquisition and preservation of digital evidence.</li>
<li>RFC 3161, Internet X.509 Public Key Infrastructure Time-Stamp Protocol (TSP). <a href="https://www.rfc-editor.org/rfc/rfc3161">IETF</a></li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>External Contractors: Who Touches Your Machines?</title>
      <link>https://trust-buster.com/blog/external-contractors-who-touches-your-machines</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/external-contractors-who-touches-your-machines</guid>
      <pubDate>Sun, 18 Oct 2026 17:00:00 GMT</pubDate>
      
      <description>External contractors are covered by documents that never reach the machine, and we want to bring the rule down to the equipment they actually touch.</description>
      <content:encoded><![CDATA[<p>It is ten past eight on a Tuesday, and the maintenance contractor&#39;s technician arrives at reception, signs the register, receives a visitor badge, retrieves the prevention plan drawn up in the spring, then crosses the workshop to the test bench he has come to adjust, and by noon he has left. Three weeks later an incident occurs on that bench, and everyone wonders who worked on it last, on which part, and whether he was still allowed to, since his qualification was due to expire at the start of the month, and it is this question, that of external contractors who touch your machines without the machine knowing it, that we want to raise here.</p>
<p><em>This article is for operators who host external companies on their equipment, and for security managers who must account for it under NIS2 or a customer&#39;s framework.</em></p>
<h2 id="a-dense-framework-around-external-contractors-that-stops-at-">A dense framework around external contractors that stops at the workshop door</h2>
<p>It would be unfair to say that external contractors are not regulated, because they are, abundantly, by texts that each have their own logic and purpose, and we take the French case as an example, as we understand it without being lawyers, while most European countries have, as far as we know, comparable rules.</p>
<p><strong>The French Labour Code</strong> organises co-activity, meaning employees of two companies working in the same place, and it provides that a joint preliminary inspection precedes the work, then that a written prevention plan is drawn up as soon as the operation represents at least 400 hours of work over twelve months, or whenever it appears on the list of hazardous work set by an order of 19 March 1993, whatever its duration in that case, and the plan identifies the interference risks between the two activities and the measures taken to prevent them.</p>
<p><strong>Electrical qualification</strong> is required for any operation on or near an electrical installation, and it is issued by the worker&#39;s employer, within the limits of the tasks entrusted and after suitable training, and since 1 October 2025 a certificate of no medical contraindication issued by the occupational physician is also required for certain operations. The host company, for its part, has every interest in checking while preparing the work that the people coming to its site are qualified for what they will do there, but that check is made on documents and before the day comes.</p>
<p><strong>The NIS2 directive</strong> requires the entities it covers to address supply chain security, including the relationships with their direct service providers (Article 21(2)(d)), as well as their access control policies and asset management (Article 21(2)(i)).</p>
<p><strong>IEC 62443-2-4</strong> describes, for its part, what an integration or maintenance provider for industrial systems must be able to guarantee to the asset owner who opens its installations to them.</p>
<p>Each of these texts produces documents, a plan, a qualification certificate, a contract clause, a policy, but none produces by itself the slightest effect on the machine, because the test bench in our example knows nothing about the prevention plan, does not know the qualification&#39;s end date and has no idea whether the technician who started it was the one expected, so control stops at the workshop door, when it does not stop at the building&#39;s.</p>
<h2 id="what-we-see-most-often">What we see most often</h2>
<p>When we walk around a site with its managers, four situations recur with a regularity that has struck us, and we describe them without seeing any particular fault.</p>
<p><strong>The shared key.</strong> The machine&#39;s key switch, or the cabinet key, circulates between the in-house team and the contractor, and people sometimes know how many copies were made, but they rarely know where they all are.</p>
<p><strong>The generic maintenance account.</strong> When the equipment has an interface, the contractor quite often uses a shared login created at commissioning and never renewed, which therefore says nothing about who typed it.</p>
<p><strong>Cascading subcontracting.</strong> The contract holder sends a subcontractor, who sometimes sends their own, and the prevention plan in principle covers everyone, but information about who is actually on site thins out at every level.</p>
<p><strong>The forgotten revocation.</strong> The contract ends, the technician changes employer or their qualification expires, and none of this translates to the field, because the field was never told anything.</p>
<p>None of these situations is, in our view, negligence in the strict sense, because they are the logical consequence of a system in which the right to intervene is written in documents, while the intervention itself happens on equipment that has no means of reading them.</p>
<h2 id="bringing-the-rule-down-to-the-equipment">Bringing the rule down to the equipment</h2>
<p>The idea we defend is simple in principle, since it consists, rather than recording the right to intervene in a document consulted after the fact, in making it a rule checked at the moment the technician stands in front of the equipment, and a qualification then becomes a dated right, attached to a person and a scope, which stops having effect on its end date without anyone having to think about it.</p>
<p>Four moments seem to us to need covering, and we built our device around them.</p>
<ol>
<li><strong>Before the work</strong>, the time slot, the equipment and the technician are entered in a rule, together with the validity date of their qualification, and this is the direct translation of the prevention plan into a language the equipment understands.</li>
<li><strong>On arrival</strong>, the technician&#39;s presence in front of the equipment is established, and their identity is verified by a cryptographic exchange with the badge or phone they carry, which is proof anchored in hardware, whereas a signature on a register remains a declaration.</li>
<li><strong>During the work</strong>, whatever is powered up, unplugged or replugged is logged, including when it falls outside the planned scope.</li>
<li><strong>Afterwards</strong>, the rule expires by itself, and the log can produce, for each technician and each piece of equipment, the list of granted and denied access.</li>
</ol>
<p>For contractor work, we have a marked preference for a mode we call Consent, in which access requires validation by an internal manager from their phone and in which nothing happens without an answer, because it fits the reality of maintenance work, which is rarely regular enough to be fully described in advance, and it leaves the decision to someone who knows the situation that day. When an emergency demands it, a break-glass procedure can force access, but the override itself is timestamped, signed and reported, because we prefer a traced exception to an invisible bypass.</p>
<h2 id="remote-or-on-site-two-distinct-topics">Remote or on site: two distinct topics</h2>
<p>Contractor management is often confused with securing remote maintenance, because the same companies and technicians are usually involved, but the two subjects are different and the solutions for one do not cover the other, since a very well designed remote-access gateway says nothing about what happens when the same technician comes on site and plugs a laptop into the controller&#39;s port, and we developed this point in <a href="/blog/remote-access-and-physical-access-two-different-problems">Remote Access and Physical Access: Two Different Problems</a>.</p>
<h2 id="what-it-does-not-replace-prevention-plan-qualification-and-l">What it does not replace: prevention plan, qualification and lockout</h2>
<p>We think it important to say what such a device does not do, so that it is not mistaken for a shortcut.</p>
<ul>
<li>It replaces neither the joint preliminary inspection, nor the prevention plan, nor the interference risk analysis, which remain obligations of the host and the external company.</li>
<li>It issues no qualification and checks no training, since it applies a validity date it was given and does not attest competence.</li>
<li>It does not stand in for any safety measure protecting people, because a lockout, known in safety practice as lockout tagout or LOTO, meaning the verified and locked de-energising of equipment before anyone works on it, remains a lockout, and we explain this at length in <a href="/blog/power-off-is-not-an-emergency-stop">Cutting the Power Is Not an Emergency Stop</a>.</li>
</ul>
<p>Finally, external contractors are not employees of the host company, but data about them is still personal data, so their presence and work should only be recorded for a clear purpose, kept for a limited time and brought to their attention, which is naturally done when the prevention plan is drawn up.</p>
<h2 id="where-to-start">Where to start</h2>
<p>If you regularly host external contractors, we would suggest starting with a modest exercise, which consists in picking the three pieces of equipment where unplanned work would have the heaviest consequences, and checking, for each one, whether you could say tomorrow who touched it last week. If the answer depends on a reception register and the team&#39;s memory, those three are a good starting point, and a deployment in plain observation, which cuts nothing, has been enough in the cases we have seen to produce the answer while changing working habits as little as possible.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>French Labour Code, Articles R4512-2 (joint preliminary inspection) and R4512-7 (written prevention plan). <a href="https://code.travail.gouv.fr/code-du-travail/r4512-7">code.travail.gouv.fr</a></li>
<li>Order of 19 March 1993 listing hazardous work requiring a written prevention plan (France).</li>
<li>French Labour Code, Articles R4544-9 and R4544-10 (electrical qualification). <a href="https://travail-industrie.com/code-du-travail/r4544-9">travail-industrie.com</a></li>
<li>INRS, <em>Habilitation électrique : foire aux questions</em>. <a href="https://www.inrs.fr/risques/electriques/habilitation-electrique-foire-aux-questions.html">inrs.fr</a></li>
<li>Decree No 2025-355 of 18 April 2025 on individual health monitoring of workers and on the qualifications under Articles R4323-56 and R4544-9 of the Labour Code (France), in force on 1 October 2025 (Article 3). <a href="https://www.legifrance.gouv.fr/jorf/id/JORFTEXT000051491350">Légifrance</a></li>
<li>Order of 26 September 2025 setting the models of certificates of no medical contraindication under Articles R4323-56 and R4544-9 of the Labour Code (France). <a href="https://content.preventionbtp.fr/pdf/droit_de_la_prevention/article-2-de-l-arrete-du-26-septembre-2025-fixant-les-modeles-d-attestation-d-absence-de-contre-indications-medicales-a-la-conduite-et-a-la-realisation-de-certaines-operations-prevues-aux-articles-r-4323-56-et-r-4544-9-du-code-du-travail.pdf">preventionbtp.fr</a></li>
<li>Directive (EU) 2022/2555 (NIS2), Article 21. <a href="https://eur-lex.europa.eu/eli/dir/2022/2555/oj">EUR-Lex</a></li>
<li>IEC 62443-2-4:2023, Security program requirements for IACS service providers. <a href="https://webstore.iec.ch/en/publication/67631">IEC Webstore</a></li>
<li>NF C18-510, Operations on electrical installations and in an electrical environment, prevention of electrical risk (France).</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>IEC 62443 and Physical Access: What the Standard Really Asks</title>
      <link>https://trust-buster.com/blog/iec-62443-and-physical-access-what-the-standard-really-asks</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/iec-62443-and-physical-access-what-the-standard-really-asks</guid>
      <pubDate>Wed, 14 Oct 2026 17:00:00 GMT</pubDate>
      
      <description>The IEC 62443 standard says little about the person at the machine, and here is where physical access appears in it, what it covers and what it does not.</description>
      <content:encoded><![CDATA[<p>The IEC 62443 standard, or more exactly the series of documents that carries this number, has become within a few years the reference that customers slip into their specifications and contracts as soon as industrial systems are involved, but it is dense, split into some fifteen documents aimed at different actors, and one often looks in vain for a direct answer to a simple question, which is what it requires regarding the person who physically walks up to a piece of equipment.</p>
<p><em>This article is for those who must meet an IEC 62443 requirement on an existing installed base, and who wonder what role physical access control plays in it.</em></p>
<h2 id="a-series-in-four-families">A series in four families</h2>
<p>Before going into detail, it helps to recall how the series is organised, because the answer to our question depends on which document one reads.</p>
<ul>
<li><strong>Parts 1</strong> set out concepts and vocabulary, that is zones, which group equipment sharing the same security requirements, conduits, which are the links between those zones, and security levels.</li>
<li><strong>Parts 2</strong> address organisations, because part 2-1 describes the asset owner&#39;s security programme and part 2-4 the requirements for integration and maintenance service providers.</li>
<li><strong>Parts 3</strong> describe the system, since 3-2 covers risk assessment and the division into zones and conduits, and 3-3 lists the system security requirements (SRs).</li>
<li><strong>Parts 4</strong> address manufacturers, 4-1 for the development lifecycle and 4-2 for the technical requirements for components (CRs).</li>
</ul>
<p>These requirements are grouped into seven foundational requirements, the same in 3-3 and 4-2, which are identification and authentication control (FR 1), use control (FR 2), system integrity (FR 3), data confidentiality (FR 4), restricted data flow (FR 5), timely response to events (FR 6) and resource availability (FR 7), and each can be met at four security levels, from SL 1 (casual or coincidental violation) to SL 4 (an attacker with extended resources).</p>
<h2 id="what-the-iec-62443-standard-says-and-where">What the IEC 62443 standard says, and where</h2>
<p>To put it simply, the series was written by and for people who think in terms of communicating systems, because it describes very precisely what a component, network or application must do to identify a user, control their rights and record their actions, but it says much less about the very common situation where the user takes no digital path at all and simply presses a button.</p>
<p>Physical access is not absent, though, and in our reading it appears in three places, in three different ways.</p>
<p><strong>In the asset owner&#39;s programme (2-1).</strong> The 2024 edition of part 2-1 reorganises the requirements into security programme elements and acknowledges, as we read it, that an installed base whose lifetime often exceeds twenty years cannot meet every technical requirement, so it leaves room for compensating measures defined by the asset owner, and that is where physical security of installations naturally belongs, alongside managing people&#39;s access.</p>
<p><strong>In component protection (4-2).</strong> The component requirements include physical tamper resistance and detection for embedded devices, hosts and network devices, so the standard accepts that the attacker may have their hands on the hardware, but it treats the problem from the viewpoint of the component being opened, not of the machine being used without authorisation.</p>
<p><strong>In compensating countermeasures.</strong> When a component cannot meet a requirement, which is the rule rather than the exception for a controller installed in 2003, the standard allows a compensating countermeasure to be provided by the surrounding system, and physical access control is one of the most natural answers to equipment that cannot identify the individual using it.</p>
<h2 id="iec-62443-3-3-requirement-by-requirement">IEC 62443-3-3, requirement by requirement</h2>
<p>It is in part 3-3 that the question becomes concrete, so take a piece of equipment with no user accounts, no log and no network, and look at what a control placed at its physical boundary can cover, and what it leaves aside.</p>
<table>
<thead>
<tr>
<th>Requirement (IEC 62443-3-3)</th>
<th>Without physical control</th>
<th>With a control at the equipment boundary</th>
</tr>
</thead>
<tbody><tr>
<td>SR 1.1 Human user identification and authentication</td>
<td>Impossible, because the equipment knows nobody</td>
<td>Compensation possible, since the person is identified before power-up</td>
</tr>
<tr>
<td>SR 2.1 Authorisation enforcement</td>
<td>Shared key or switch</td>
<td>Dated rule, with the right person, on the right equipment, at the right time</td>
</tr>
<tr>
<td>SR 2.8 Auditable events</td>
<td>None</td>
<td>Grants, denials and bypass attempts logged</td>
</tr>
<tr>
<td>SR 2.11 Timestamps</td>
<td>None</td>
<td>Every entry timestamped</td>
</tr>
<tr>
<td>SR 2.12 Non-repudiation</td>
<td>None</td>
<td>Signed entries, attributed to an identity</td>
</tr>
<tr>
<td>SR 6.1 Audit log accessibility</td>
<td>Not applicable</td>
<td>Logs retrievable and exportable to a SIEM, the tool that centralises security logs</td>
</tr>
<tr>
<td>SR 7.8 Control system component inventory</td>
<td>Declarative, often incomplete</td>
<td>Field inventory of protected equipment and of transmitters present</td>
</tr>
<tr>
<td>SR 3.x System integrity</td>
<td>Not addressed</td>
<td>Not addressed, because we do not touch the equipment&#39;s logic</td>
</tr>
<tr>
<td>SR 5.1 Network segmentation</td>
<td>Out of scope</td>
<td>Out of scope, since the equipment stays off the network</td>
</tr>
</tbody></table>
<p>Two observations seem important to us. The first is that the middle column is not a caricature, because it is the actual state of a large part of the installed base, and it is what, as far as we know, auditors who ask the question find. The second is that the right-hand column does not cover everything, and it would be misleading to claim otherwise, because a boundary control does not make the equipment more intact, but it makes its use attributable and verifiable.</p>
<h2 id="what-the-auditor-will-ask-for">What the auditor will ask for</h2>
<p>In practice, an assessment against the IEC 62443 standard is not satisfied with a declaration of coverage, because it asks for evidence, and that is often where compensating measures fail, since an unevenly kept paper register proves nothing, and an unapplied procedure proves even less.</p>
<p>What gives a compensating measure its value is that it produces its own evidence, so a grant, a denial or a bypass attempt must leave a dated, attributed, tamper-evident trace without anyone having to remember to record it, and that is the property we sought by signing each log entry as it is produced, rather than relying on human discipline for traceability.</p>
<h2 id="what-we-do-not-claim">What we do not claim</h2>
<p>The standard lends itself easily to marketing shortcuts, and we would rather be precise about our own position.</p>
<ul>
<li>Our architecture follows the design principles of IEC 62443-4-2, targeting security level SL-C 2, but no third-party assessment has been carried out to date, so we claim no certified compliance.</li>
<li>No product makes a system IEC 62443 compliant, because compliance belongs to a system, assessed in its context, with its zones, conduits and security programme, and a product brings capabilities and evidence, which is already a lot, but it is not the same thing.</li>
</ul>
<h2 id="where-to-start">Where to start</h2>
<p>If you must meet a requirement of the IEC 62443 standard on an existing installed base, we would suggest starting with a deliberately simple question, which is, for each critical piece of equipment in your zones, who can start it and how you would know tomorrow, because where the answer is &quot;anyone&quot; and &quot;we would not&quot;, you have found the place where a physical compensating measure has the most value, and where it will be easiest to defend in front of an assessor.</p>
<p>For the layered model underlying the standard, our Purdue series starts with <a href="/blog/purdue-diligence">Purdue Diligence</a>, and the link with the NIS2 directive is covered in <a href="/blog/how-to-navigate-nis2-compliance-for-your-business">How to Navigate NIS2 Compliance for Your Business</a>.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>IEC 62443-2-1:2024, Security for industrial automation and control systems, Part 2-1: Security program requirements for IACS asset owners. <a href="https://webstore.iec.ch/en/publication/62883">IEC Webstore</a></li>
<li>IEC 62443-2-4:2023, Security program requirements for IACS service providers. <a href="https://webstore.iec.ch/en/publication/67631">IEC Webstore</a></li>
<li>IEC 62443-3-2:2020, Security risk assessment for system design.</li>
<li>IEC 62443-3-3:2013, System security requirements and security levels.</li>
<li>IEC 62443-4-2:2019, Technical security requirements for IACS components.</li>
<li>ISA, overview of the ISA/IEC 62443 series. <a href="https://www.isa.org/standards-and-publications/isa-standards/isa-iec-62443-series-of-standards">isa.org</a></li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>Cutting the Power Is Not an Emergency Stop</title>
      <link>https://trust-buster.com/blog/power-off-is-not-an-emergency-stop</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/power-off-is-not-an-emergency-stop</guid>
      <pubDate>Sun, 11 Oct 2026 17:00:00 GMT</pubDate>
      
      <description>Cutting the power to a machine is not an emergency stop, nor a lockout, nor a safety function, and here is what that changes in how we designed the Cube.</description>
      <content:encoded><![CDATA[<p>When we present Trust-Buster to people we know who work in this sector, the first objection is almost never technical in the sense IT people mean it, because it fits in one sentence, often asked with some suspicion, which is &quot;you can cut the power to a machine, so you are doing machine safety, so show me your performance level&quot;. The question is fair, and we would rather answer it plainly, because no, we do not do machine safety in the functional-safety sense, cutting the power is not an emergency stop, and that is precisely why our device is designed the way it is.</p>
<p><em>This article is for operators, security managers and electricians wondering where electrical access control sits next to the personnel-protection devices they already know.</em></p>
<h2 id="two-words-two-trades">Two words, two trades</h2>
<p>English has two words, safety and security, and French has sûreté and sécurité, but French usage mixes them, and the personnel-protection standards themselves are titled &quot;safety of machinery&quot;, so we fix our vocabulary for the rest of this article. <strong>Safety</strong> protects people and property against failures, errors and the hazards of the machine itself, with an emergency stop, an interlocked guard, a light curtain or a safety relay. <strong>Security</strong>, in the sense we use it, protects against unauthorised use, malice and bypass, and it answers the question of who is allowed to start this equipment, when, and how to prove it.</p>
<p>Both disciplines have their standards, methods and owners. Machine safety relies on a risk assessment and on safety functions whose reliability is demonstrated, typically under EN ISO 13849-1 (performance levels, PL) or IEC 62061 (safety integrity levels, SIL), while industrial system security relies on other frameworks, first among them the IEC 62443 series. The two increasingly meet, and the new European Machinery Regulation, applicable from 20 January 2027, makes the meeting mandatory by requiring that safety functions cannot be compromised by malicious action, but meeting is not merging.</p>
<h2 id="what-our-device-is-not-starting-with-the-emergency-stop">What our device is not, starting with the emergency stop</h2>
<p>To avoid any misunderstanding, we would rather list what we do not claim than leave the reader to guess.</p>
<p><strong>It is not an emergency stop.</strong> An emergency stop meets precise requirements (EN ISO 13850), because it is accessible, it overrides every other command, it acts according to a stop category chosen by the risk assessment, and it must be deliberately reset. Stop categories 0, 1 and 2 are defined by EN 60204-1, and EN ISO 13850 refers to these categories for the emergency stop. A socket driven by an access rule can cut the power and restore it, but it meets none of these conditions, and it must never be presented as one.</p>
<p><strong>It is not lockout.</strong> Electrical lockout, described in France by the NF C18-510 standard and elsewhere by equivalent lockout and tagout rules, requires a reliable separation from the energy source, a physical lock on the isolating device, identification of the installation and a verification that no voltage is present, whereas a remotely controlled relay, however reliable, offers neither visible separation nor padlocking, so nobody should ever work on a machine because the Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), has switched it off.</p>
<p><strong>It is not a safety function within the meaning of safety standards.</strong> Our effector, that is the module that opens or closes the power supply, is neither assessed nor certified under EN ISO 13849-1 or IEC 62061, and we never insert it into an existing safety chain, because it sits upstream, on the equipment&#39;s power supply, next to the protective devices and never in their place.</p>
<p>These three limits we write into our contracts and on our product pages, because in our view a sales pitch that left any doubt on this point would do more harm than good.</p>
<h2 id="why-cutting-power-can-create-a-hazard">Why cutting power can create a hazard</h2>
<p>It is tempting to think that it is enough to cut the power for a machine to be safe, and it often is, which is what makes the power supply interesting as a control point, but not always, and the exceptions are exactly the ones a safety manager cares about.</p>
<p><strong>Uncontrolled stop.</strong> Abruptly removing energy from a running machine is a category 0 stop in the sense of EN 60204-1, and for a bench drill it does not matter, but for a machine with an axis holding a load, an undersized fail-safe brake, a workpiece held by vacuum or a thermal process that must run to completion, the interruption may create the very risk one wanted to avoid.</p>
<p><strong>Restart when power returns.</strong> EN 60204-1 requires that a machine does not restart dangerously on its own after a supply interruption followed by restoration, and EN ISO 14118 deals more broadly with preventing unexpected start-up. Many machines comply, but not all do, especially in older equipment, laboratories and shared workshops, so a device that cuts and then restores power must be designed never to become the cause of an unexpected restart itself.</p>
<p><strong>The electrical installation.</strong> Inserting a switching device on a cabinet outgoing circuit means modifying an installation, with everything that implies in terms of calculations, selectivity, protection of persons and updated documentation, and it is not a trivial act, nor one we perform alone.</p>
<h2 id="what-we-drew-from-this-in-the-design">What we drew from this in the design</h2>
<p>These findings did not lead us to give up on power-supply control, but to constrain it, and four principles follow, which we present as deliberate choices rather than general truths.</p>
<p><strong>Authorise start-up rather than interrupt.</strong> By default, our device governs the authorisation to power up, not the stop, so equipment that is already running and falls outside its rule triggers an alert and not a cut, unless it appears explicitly on a list of equipment that may be cut while running, drawn up and signed with the operator after examining what an interruption would cause. That list holds a lamp, a microscope, a glue gun or a soldering iron, in short what is handheld, has a standby mode and is used to being switched off, and we know very well that in a workshop things are not always switched off properly, but if it is your installation you may choose to leave the microscope powered until someone comes and presses its button.</p>
<p><strong>Come back powered off.</strong> After a mains failure, our enforcement points come back with the power off, and returning to service is an authorised act, never a side effect, because we preferred that a power return leaves the machine off, even at the cost of one more action, rather than hand control back to equipment whose restart behaviour we do not know.</p>
<p><strong>Observe before arming.</strong> Every deployment starts in Observation mode, which detects, logs and alerts without ever cutting, and during that period the real behaviour of the equipment is measured, and the decision whether to go further is taken one machine at a time, as we describe in <a href="/blog/what-the-cube-knows-about-your-machines">What the Cube Knows About Your Machines</a>. Many sites never do, and that is a perfectly honourable outcome.</p>
<p><strong>Stay on the socket or a dedicated circuit.</strong> Our operational effector currently acts on single-phase 230 V supplies, as a controlled socket or a DIN-rail module, while a hard-wired three-phase machine requires a study, a qualified electrician and an update of the installation file, so we handle it case by case, and we do not sell it as an off-the-shelf product.</p>
<h2 id="what-about-substantial-modification">What about substantial modification?</h2>
<p>Regulation (EU) 2023/1230 defines a substantial modification as a change not foreseen by the manufacturer, made after the machine was placed on the market, that creates a new hazard or increases an existing risk to the point of requiring new protective measures, and whoever makes it becomes, under the Regulation, the manufacturer of the modified machine.</p>
<p>A device placed upstream of the power supply, which touches neither the control system nor the safety functions of the machine, is not in principle meant to fall into that category, and this is one of the reasons we stay at the boundary of the equipment rather than entering it, but we would not turn &quot;in principle&quot; into a certainty, because the answer depends on the machine, its operating mode and what a power cut does to it. It is a question to raise during the risk assessment, with the person responsible for the machine&#39;s conformity, and not one to settle in a brochure.</p>
<h2 id="complementary-rather-than-competing">Complementary rather than competing</h2>
<p>In the end it seems fairer to describe our device as a layer added to existing protections, and one that actually helps them, because a lockout remains a lockout, but the Cube&#39;s log can show that no authorised power-up occurred during it, an emergency stop remains an emergency stop, but knowing who started the machine ten minutes before someone had to press it sheds light on the incident analysis, and an electrical qualification is still issued by the employer, but its expiry date can become a rule instead of a line in a spreadsheet.</p>
<p>That is the fit we are looking for, and it is also why we insist that the decision to arm a piece of equipment stays with the operator, with a named safety referent on their side. Safety remains the business of those whose trade it is, and being able to cut the power to a circuit changes nothing about that, because our role is to give them what they lacked until now, namely a verifiable answer to the question &quot;who powered this machine, and were they allowed to&quot;.</p>
<h2 id="sources">Sources</h2>
<ul>
<li>Regulation (EU) 2023/1230 of the European Parliament and of the Council of 14 June 2023 on machinery, Article 3(16) (substantial modification) and Article 18. <a href="https://eur-lex.europa.eu/eli/reg/2023/1230/oj">EUR-Lex</a></li>
<li>EN 60204-1:2018, Safety of machinery, electrical equipment of machines, part 1: general requirements (stop categories, protection against supply restoration).</li>
<li>EN ISO 13850:2015, Safety of machinery, emergency stop function, principles for design.</li>
<li>EN ISO 14118:2017, Safety of machinery, prevention of unexpected start-up.</li>
<li>EN ISO 13849-1:2023, Safety of machinery, safety-related parts of control systems, part 1: general principles for design.</li>
<li>IEC 62061:2021, Safety of machinery, functional safety of safety-related control systems.</li>
<li>NF C18-510, Operations on electrical installations and in an electrical environment, prevention of electrical risk (France).</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>What the Cube Knows About Your Machines</title>
      <link>https://trust-buster.com/blog/what-the-cube-knows-about-your-machines</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/what-the-cube-knows-about-your-machines</guid>
      <pubDate>Wed, 16 Sep 2026 09:54:37 GMT</pubDate>
      
      <description>A smart meter measures what a machine consumes but not who was in the room, and here is what the Cube knows by crossing measurement, presence and sensors.</description>
      <content:encoded><![CDATA[<p>A smart meter costs around thirty euros and measures quite accurately what a machine consumes, but it does not know <a href="/blog/who-is-in-the-room">who was in the room</a>. The Cube is our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), and what the Cube knows about your machines comes from correlating consumption, enrolled presences and the sensors you already have to produce what none of the three sources provides on its own, namely an attributed usage log, which serves to recharge, to spot use outside scheduled hours or to prove later who started what. It does so in Observation mode, that is by watching and recording without ever cutting anything off.</p>
<h2 id="measurement-presence-correlation-activity-reports-as-things-">Measurement, presence, correlation: activity reports as things happen</h2>
<p>An off-the-shelf module clipped onto a cable measures consumption with sufficient precision, but it knows nothing of who was in the room, nor of what it was normal to consume that day, nor of what ought to trigger an alert.</p>
<p>What the Cube produces is of a different nature, because it is an attributed activity report, reasoned alerts and a signed log usable in an audit, and sensors are its raw material, not its product. That is also why we do not seek to manufacture sensors, but to read them in order to complete the Cube&#39;s decision, which consists in logging an authorised action, alerting on inappropriate use or an intrusion, and as a last resort only, if the operator has so decided, placing equipment in a protected state through a cut-off defined according to safety and security rules, as we explain in Cutting the Power Is Not an Emergency Stop.</p>
<h2 id="three-input-sources">Three input sources</h2>
<p><strong>The effectors</strong>, first. These are the modules we fit ourselves on the equipment&#39;s power supply, and they have a useful particularity, since they measure and control at the same point, which makes it possible to move from observation to action without reinstalling anything on the day the operator decides to.</p>
<p><strong>Presence</strong>, next, established by the Cube from the enrolled badges, phones or connected objects that enter and leave the premises and indicate the presence of known users.</p>
<p><strong>And your sensors</strong>, if needed. This is the point we long underestimated in our own communication, because the Cube knows how to read sources it did not install, such as temperature and humidity probes, existing energy meters polled over Modbus, which is a common industrial protocol, door contacts, flow meters or points from a building management system (BMS) already in place.</p>
<p>On an instrumented site, a good share of the raw material is already there, but it has never been cross-referenced with anything else, because a temperature reading that lives in a supervision dashboard and an entry register that lives in a binder have never met.</p>
<img loading="lazy" src="https://trust-buster.com/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__adec2e60-fbed-4ae3-bcf2-10d0bf8673ab_2048x2048-960.webp" alt="What the Cube knows comes from three sources, an effector on the power supply, a badge and sensors, and a log comes out" data-image-fit="fit" srcset="/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__adec2e60-fbed-4ae3-bcf2-10d0bf8673ab_2048x2048-480.webp 480w, /media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__adec2e60-fbed-4ae3-bcf2-10d0bf8673ab_2048x2048-960.webp 960w, /media/ce254880-8bbf-45bf-baa3-8bfd55ed258a__adec2e60-fbed-4ae3-bcf2-10d0bf8673ab_2048x2048.webp 1376w" sizes="(min-width: 1024px) 1024px, 100vw" width="1376" height="768">

<h2 id="why-control-and-measurement-at-the-same-point">Why control and measurement at the same point</h2>
<p>A four-channel DIN-rail module, installed inside the electrical cabinet, brings four things at once.</p>
<p><strong>Four circuits for a single device.</strong> Measurement and control on four outgoing feeders, with a single connection point and a single device to enrol, and the saving is real across an estate.</p>
<p><strong>Measurement and cut-off inseparable.</strong> The same electronics observe and act on the same channel, so there is no configuration in which one circuit is measured and its neighbour controlled, and the report and the action concern the same object, by construction.</p>
<p><strong>The device is out of reach.</strong> This is the point we insist on, because a module left in plain sight on a socket can be unplugged, reset or carried off, while the same module behind the closed door of a cabinet inherits the cabinet&#39;s protection. The residual risk we document elsewhere, namely that a factory reset restores the module&#39;s original behaviour, is much reduced when access to the module already requires opening a locked cover.</p>
<p><strong>And the machine is not touched.</strong> The cabinet in which we fit the module is the distribution cabinet, not the machine&#39;s, so we stay at the boundary, on the feeder that powers the equipment and not on the equipment.</p>
<h2 id="what-a-power-curve-tells">What a power curve tells</h2>
<p>A machine does not consume uniformly, and the levels of its load curve are readable.</p>
<p><strong>When fully stopped</strong>, consumption is zero or near zero, and it is the only unambiguous state.</p>
<p><strong>On standby</strong>, it is low but not zero, because of the control electronics, the display or the cabinet ventilation. It is a stable level, specific to each machine, and it is the one most often mistaken for being off, because a &quot;switched-off&quot; machine that permanently draws forty watts is a machine that is on.</p>
<p><strong>At start-up</strong>, a motor draws a current markedly higher than its rated current for a short moment, and that peak is a signature, because it dates the start-up to the second provided the measurement is frequent enough, and because it distinguishes a genuine start from a wake from standby.</p>
<p><strong>Running idle</strong>, the machine turns without working, and <strong>under load</strong> it consumes more, with a gap that is generally clear-cut.</p>
<p>From these levels follow four pieces of information usable without interpretation, which are the number of start-ups, the time powered on, the time under load and the energy consumed, and the difference between the two durations is often the most telling figure, because a machine powered on for eight hours and under load for forty minutes is not the same management object depending on which one you look at.</p>
<img loading="lazy" src="https://trust-buster.com/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__108c9a17-1733-4cfc-9f68-406791b33b8d_2048x2048-960.webp" alt="Power curve of a machine with its levels, standby, start-up peak, running idle, under load and back to stop" data-image-fit="fill" srcset="/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__108c9a17-1733-4cfc-9f68-406791b33b8d_2048x2048-480.webp 480w, /media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__108c9a17-1733-4cfc-9f68-406791b33b8d_2048x2048-960.webp 960w, /media/ce254880-8bbf-45bf-baa3-8bfd55ed258a__108c9a17-1733-4cfc-9f68-406791b33b8d_2048x2048.webp 1376w" sizes="(min-width: 1024px) 1024px, 100vw" width="1376" height="768">

<h2 id="what-the-cube-knows-once-it-crosses-the-sources">What the Cube knows once it crosses the sources</h2>
<p>Four situations emerge, and each calls for a different reading.</p>
<p><strong>Attributed use.</strong> The machine is running and a single enrolled person is present throughout the period, which is the nominal case and produces a usable log line, with who, which machine, from when to when and how much energy.</p>
<p><strong>Unattributed use.</strong> The machine is running and nobody is detected. Three explanations are possible and all three must be kept in mind, a scheduled cycle that requires nobody, a person present without an enrolled identifier or a remote start, so this is not an alert in itself but a line that calls for an explanation, and it is more frequent than one might imagine.</p>
<p><strong>Presence without use.</strong> Someone came and the machine did not start, which is commonplace but useful for distinguishing a visit from an intervention.</p>
<p><strong>Change of state during a presence.</strong> This is the most interesting case, because nobody sees a thermostat being turned, but if the heating elements start drawing more while an identified person was on site, and nothing in the outside conditions explains it, a setpoint has probably moved.</p>
<p>And this is precisely where your existing sensors make the difference, because with consumption alone the inference is indirect and fragile, while with an indoor temperature probe as well, which you probably already have, the reasoning becomes clear-cut, since the setpoint temperature has evidently changed, at such a time, and here is who was there. We move from a hypothesis to an observation.</p>
<img loading="lazy" src="https://trust-buster.com/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__7b8e0468-ff02-482c-ade5-cfba19fde847_2048x2048-960.webp" alt="Observed consumption compared with the expected baseline, with a deviation highlighted over one period" data-image-fit="fill" srcset="/media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__7b8e0468-ff02-482c-ade5-cfba19fde847_2048x2048-480.webp 480w, /media/_w/ce254880-8bbf-45bf-baa3-8bfd55ed258a__7b8e0468-ff02-482c-ade5-cfba19fde847_2048x2048-960.webp 960w, /media/ce254880-8bbf-45bf-baa3-8bfd55ed258a__7b8e0468-ff02-482c-ade5-cfba19fde847_2048x2048.webp 1376w" sizes="(min-width: 1024px) 1024px, 100vw" width="1376" height="768">

<h3 id="establishing-baselines">Establishing baselines</h3>
<p>High consumption means nothing in itself, because a heater consumes more on a cold night, a pump when the level drops and a compressor in summer, so the usable signal is the gap between the observed consumption and the consumption expected under those particular conditions.</p>
<p>This requires measuring first the normal behaviour of each piece of equipment over several weeks, with its seasonal and weekly variations, and that is what Observation mode does, and that is why it takes time, because what the Cube knows about a piece of equipment after three days is not worth much.</p>
<p>A system that alerts on an absolute threshold is likely to produce false alerts all winter and to end up unplugged before spring, and a baseline that incorporates outdoor temperature is markedly finer than a calendar-based one, which is one more reason to read your sensors.</p>
<h3 id="the-inference-scale">The inference scale</h3>
<p>Here is the grid we are keen to set out explicitly, because this is where the gap lies between an honest use and an abusive use of this data.</p>
<table>
<thead>
<tr>
<th>Level</th>
<th>Example</th>
<th>Status</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Measured</strong></td>
<td>1,240 W at 14:32, badge B present from 14:10 to 15:05, 19.4 °C indoors</td>
<td>Raw data, not open to dispute</td>
</tr>
<tr>
<td><strong>Established</strong></td>
<td>The machine was powered on from 14:28 to 15:02, of which 22 minutes under load</td>
<td>Direct deduction from the levels</td>
</tr>
<tr>
<td><strong>Inferred, high confidence</strong></td>
<td>The machine was used, not merely switched on</td>
<td>Relies on a load threshold calibrated per piece of equipment</td>
</tr>
<tr>
<td><strong>Inferred, moderate confidence</strong></td>
<td>A setpoint was changed during this presence</td>
<td>Relies on a deviation from the baseline, and becomes high confidence if a temperature sensor confirms it</td>
</tr>
<tr>
<td><strong>Not inferable</strong></td>
<td>What the person did, what was produced, whether it was authorised, who exactly acted if several people were present</td>
<td>Never to be presented as a conclusion</td>
</tr>
</tbody></table>
<p>The last row is the one that matters to us, because it is the one a vendor is tempted to cross, and a report that states &quot;so-and-so changed the thermostat&quot; says more than the data allows. What it can honestly write is that &quot;heating consumption deviated from the baseline by 34% at 14:40, that the indoor setpoint temperature rose by 2 °C over the same period, and that the people present were A and B&quot;.</p>
<p>The difference is not cosmetic, because the first wording does not survive a challenge, while the second one does.</p>
<h2 id="what-the-cube-will-never-say">What the Cube will never say</h2>
<p>Everything the Cube knows fits in the grid above, so there remain things it will not say.</p>
<p><strong>Power says the machine ran, not what it did</strong>, because no electrical measurement distinguishes a compliant part from a reject.</p>
<p><strong>A shared circuit blurs attribution.</strong> Three machines on the same feeder, and you measure their sum, which is, incidentally, the concrete argument in favour of a four-channel module rather than one module per distribution board.</p>
<p><strong>Two people present, no individual attribution.</strong> This is structural, and it is probably a good thing, because it makes the system poorly suited to individual performance monitoring, and we do not seek to make it suited to that.</p>
<p><strong>A badge proves a badge.</strong> A lent badge produces a false attribution, dated and signed.</p>
<p><strong>And the baseline can be fooled.</strong> Someone who knows the system can change a setpoint gradually, over several weeks, so that each deviation stays below the threshold, and we do not claim to detect that.</p>
<h2 id="the-legal-question-which-is-not-the-same-as-before">The legal question, which is not the same as before</h2>
<p>Measuring that a machine ran is technical data, while measuring that a named person was present is personal data within the meaning of the GDPR, and combining the two to deduce what a person did and for how long falls under the monitoring of employee activity, which is not the same regime. We have detailed what that implies in Detecting Presence Without Monitoring Employees, and what the Cube knows about a person stops at their presence.</p>
<p>What makes such processing legitimate is its purpose, because recharging the use of a shared workshop is a clear and proportionate purpose, alerting on operation outside scheduled hours is one as well, while drawing up an individual productivity ranking is not, and a system that allowed it should not have been designed that way.</p>
<p>Three consequences follow, drawn at the design stage rather than in a usage policy.</p>
<p><strong>Aggregation is the default mode.</strong> Reports are produced per machine, per period and per organisation, and named attribution appears only on a defined trigger, such as use outside scheduled hours, an anomaly or a reasoned audit request.</p>
<p><strong>Retention periods are distinct.</strong> That of named correlations is short, while that of consumption data, which has nothing personal about it once detached from presence, follows your management needs.</p>
<p><strong>And the system announces itself.</strong> Installing an observation system in premises where employees work requires, in France, informing and consulting the CSE, the works council, before installation and not after.</p>
<h2 id="why-start-there">Why start there</h2>
<p>For a large number of operators, this mode is already a sufficient result, without going as far as on and off access control, because they never needed a machine to stop, but they needed to know who had started it, and to be able to prove it eighteen months later.</p>
<p>This has a pleasant practical consequence, because a deployment in observation cannot, as long as nothing is armed, stop your production, so the question &quot;and what if your box gets it wrong?&quot; loses most of its weight, and with it a good part of the difficulty in getting a pilot accepted.</p>
<p>What you obtain nonetheless is an attributed usage log, detection of operation outside scheduled hours, a basis for recharging and the actual event rate per machine on your site, that is the data that allows you to decide, later and with full knowledge of the facts, whether it is worth arming anything at all.</p>
<p>And if you one day decide to do so, nothing needs reinstalling, because it is the same device, in the same cabinet, that was already measuring, and what the Cube knows already about the machine then allows that decision to be taken without haste.</p>
]]></content:encoded>
    </item>
    <item>
      <title>Remote Access and Physical Access: Two Different Problems</title>
      <link>https://trust-buster.com/blog/remote-access-and-physical-access-two-different-problems</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/remote-access-and-physical-access-two-different-problems</guid>
      <pubDate>Sat, 12 Sep 2026 12:16:22 GMT</pubDate>
      
      <description>Securing remote access to industrial equipment and controlling physical access to it are two different problems, and neither one covers the other.</description>
      <content:encoded><![CDATA[<p>Same technician, same effects, but two separate forms of management, because securing remote access to a piece of industrial equipment and controlling physical access to that same equipment are two different problems, and neither one covers the other. A bastion, that is the jump server through which sensitive connections are routed so that they can be controlled and recorded, protects the session of the technician connecting from another site, but it says nothing about that same technician standing in front of the cabinet the following Tuesday.</p>
<p><em>These two subjects are regularly confused, and the confusion is costly because people buy one believing they are dealing with the other, so here is what each one covers as far as we know.</em></p>
<h2 id="what-secure-remote-access-covers">What secure remote access covers</h2>
<p>Let us begin by describing it properly, because it is a demanding discipline and one the model describes well. A serious remote access set-up (bastion, application gateway, access broker) handles a complete and coherent chain.</p>
<p><strong>The requester&#39;s identity</strong> is federated with the operator&#39;s directory or the service provider&#39;s, with multi-factor authentication, so we know who is connecting, and the proof is cryptographic.</p>
<p><strong>Access is granted on demand and for a bounded period.</strong> The manufacturer&#39;s account does not exist permanently, since it is activated for one remote maintenance intervention, on a designated piece of equipment, during a defined window.</p>
<p><strong>The session is supervised.</strong> What was done is recorded in full, sometimes with the ability to take it over live or to cut it off, and this is the most useful thing privileged access management brings, namely a usable trace.</p>
<p><strong>The path is controlled.</strong> The connection does not cross the architecture end to end, because it stops in the demilitarised zone (DMZ) at level 3.5, where a protocol break takes place, and a jump host takes over towards the target system. Outbound-only connectivity, recommended by ANSSI and CISA alike, completes the set-up.</p>
<p><strong>And the scope is restricted.</strong> The technician only reaches the equipment they have been mandated for, not the workshop network.</p>
<p>This is solid, and an industrial site that does not have it has a more urgent problem to solve than the one this article is about.</p>
<h2 id="what-the-bastion-does-not-cover">What the bastion does not cover</h2>
<p>Now let us take the same technician, in a different situation.</p>
<p>They take a plane and arrive on site, they open the cabinet, they plug their laptop into the programming port on the front panel, and they do exactly what they would have done remotely, sometimes even with the same software, with the same rights on the PLC and the same ability to reprogramme it.</p>
<p>The bastion saw nothing, because it was not on the path.</p>
<p>This is not a flaw in the remote access product, since it does exactly what it was designed for, and blaming it would be absurd, so it is a question of scope. Privileged access management covers the manufacturer&#39;s remote access, but their physical access, same person and same service, is most often covered by no technical measure at all, even when a normative requirement addresses it.</p>
<p>And this is not a laboratory hypothesis, since on most of the sites we know physical intervention remains the main path by which a level 0 or level 1 device gets modified, because such equipment cannot be configured any other way.</p>
<h2 id="remote-access-and-physical-access-line-by-line">Remote access and physical access, line by line</h2>
<table>
<thead>
<tr>
<td></td>
<th>Secure remote access</th>
<th>Controlled physical access</th>
</tr>
</thead>
<tbody><tr>
<td><strong>The subject</strong></td>
<td>A session opened from outside</td>
<td>A person present in front of the equipment</td>
</tr>
<tr>
<td><strong>The path</strong></td>
<td>A network conduit, drawn in the architecture</td>
<td>A door, drawn nowhere</td>
</tr>
<tr>
<td><strong>The enforcement point</strong></td>
<td>Bastion, gateway, protocol break at level 3.5</td>
<td>Actuator on the power supply or the physical interface</td>
</tr>
<tr>
<td><strong>The proof of identity</strong></td>
<td>Federation with a directory, multiple factors</td>
<td>Cryptographic challenge on a secure element carried by the person</td>
</tr>
<tr>
<td><strong>The proof of use</strong></td>
<td>Session recording</td>
<td>Signed log of power-ups, cable connections and refusals</td>
</tr>
<tr>
<td><strong>The network dependency</strong></td>
<td>Total, by nature</td>
<td>None, because the decision is local</td>
</tr>
<tr>
<td><strong>What it sees of the technician on site</strong></td>
<td>Nothing</td>
<td>What goes through the power supply or the controlled interface</td>
</tr>
<tr>
<td><strong>What it sees of the remote technician</strong></td>
<td>Everything</td>
<td>Nothing</td>
</tr>
</tbody></table>
<p>The two set-ups are complementary and disjoint, so when you secure one you do not gain the other as a bonus effect, because a technician standing in front of the machine does not open a remote session to reach it, but uses the front-panel port, and often other access points as well, which makes the problem worse rather than solving it.</p>
<h2 id="why-the-confusion-exists">Why the confusion exists</h2>
<p>Three reasons, two of which are legitimate.</p>
<p><strong>The vocabulary is identical.</strong> Access control, identity, authorisation, privilege, session, revocation, both fields use the same words for different objects, and a buyer reading two brochures sees the same thing twice.</p>
<p><strong>The Purdue model does not distinguish between them</strong>, since it does not describe the person at all. As we saw in the third article of this series, <a href="/blog/long-purdue">Long Purdue</a>, only one of the two paths appears on the diagram, so the other has no name, and what has no name cannot be compared.</p>
<p><strong>And there is a less admissible reason</strong>, because it is commercially tempting to let people believe that one product covers both. Nobody is really lying, but a feature table that ticks &quot;access control for industrial equipment&quot; without specifying which kind leaves the buyer to fill in the blank on their own.</p>
<h2 id="the-case-that-brings-them-together">The case that brings them together</h2>
<p>There is one situation in which the two set-ups meet, and it deserves to be laid out because it is common.</p>
<p>The manufacturer&#39;s technician intervenes remotely for a diagnosis, concludes that a component needs replacing, and comes on site the following week, so same intervention, same work order, same person, and two paths.</p>
<p>An operator who governs this case well ought to be able to answer a simple question, which is whether the work order is the same object in both systems. In other words, are the authorised window, the designated equipment and the verified clearance consistent between the remote session and the site visit, or are they two unrelated authorisations?</p>
<p>In almost every case we know of, they are two unrelated authorisations, because the bastion knows a technical account and a time window, while the site reception knows a name and a visitor badge, and nothing links the two until an incident occurs. We described what a site visit implies on the operator&#39;s side in External Contractors: Who Touches Your Machines?.</p>
<p>That is probably where the real work of the coming years lies, and it will be done neither by the remote access players alone, nor by us alone.</p>
<h2 id="where-we-stand">Where we stand</h2>
<p>We do not do secure remote access and we have no intention of doing so, because it is a demanding discipline, occupied by competent players, and the boundary between levels 3 and 4 in all likelihood remains the control that weighs most in your architecture. In small shops without a bastion, a satellite can condition the provider&#39;s existing access without replacing it.</p>
<p>We do the other path, that of the person who is present, on equipment that has no screen, no identity and no network, with a decision rendered locally even when nothing else is responding, and it is because these two paths, remote access and physical access, remain two different problems that we prefer to say clearly which of the two we address.</p>
]]></content:encoded>
    </item>
    <item>
      <title>Long Purdue: The Link the Purdue Model Is Missing</title>
      <link>https://trust-buster.com/blog/long-purdue</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/long-purdue</guid>
      <pubDate>Fri, 11 Sep 2026 12:37:28 GMT</pubDate>
      
      <description>A technician plugged into a PLC has crossed no firewall and exists at no level of the model, and that Purdue missing link, the person, is our subject.</description>
      <content:encoded><![CDATA[<p>If we are asked where our system sits in the Purdue model, the answer is simple, because it sits nowhere, and it was created precisely to fill that gap, since the model describes systems talking to systems and is missing one link, the person standing in front of the machine, which we call in this article the Purdue missing link, for want of a better name.</p>
<h2 id="the-purdue-model-has-no-level-for-the-person">The Purdue model has no level for the person</h2>
<p>A manufacturer&#39;s technician who opens a cabinet and plugs a laptop into the programming port of a PLC, a programmable logic controller, has crossed no firewall, no demilitarised zone and no segmentation rule, and yet they have walked straight into level 1, the level of basic control, while their link meets, in our view, what IEC 62443-3-2 calls a communication path between zones, even though no architecture diagram draws it.</p>
<h2 id="an-ordinary-tuesday-morning">An ordinary Tuesday morning</h2>
<p>A manufacturer&#39;s technician arrives on site at nine o&#39;clock, reports to reception where he is handed a visitor badge, and a maintenance manager escorts him to workshop B, then he opens the cabinet of a drive, the electronic unit that sets a motor&#39;s speed, with the triangle key lying around in the toolbox, plugs his laptop into the front programming port, opens his manufacturer software, changes two acceleration ramp parameters, uploads, closes the cabinet and leaves at half past eleven.</p>
<p>Let us retrace his route from the point of view of the security architecture, because he crossed no firewall, he did not pass through the level 3.5 demilitarised zone nor open a session on a bastion host, no microsegmentation rule applied to him, no unidirectional gateway stood in his way and no detection system saw an unusual packet go by, since his laptop and the PLC talked one-to-one over a serial port, outside any network, so he walked straight into level 1 through a door and not a port.</p>
<p>And almost nobody today could answer three elementary questions:</p>
<ul>
<li>exactly which parameters were changed,</li>
<li>on what contractual basis,</li>
<li>and whether it was really he who changed them.</li>
</ul>
<h2 id="why-the-model-does-not-see-it">Why the model does not see it</h2>
<p>This is not an oversight on the part of the Purdue consortium, and it would be unfair to present it that way, because the model was designed to organise the data flows of a computerised plant and each of its levels connects one system to another system, since level 0 is the physical process with its sensors and actuators, level 1 is basic control with its PLCs and controllers, level 2 supervises, level 3 manages operations, and so on, as we <a href="/blog/purdue-diligence">detailed in the first article</a>.</p>
<p>Nowhere is there a level for the human present in the field, because that was not the question being put to the model in 1992, and the operator appears in it implicitly behind a level 2 interface as the user of a system, while the maintainer, the Purdue missing link we are talking about, does not appear at all, because they do not use a system, they open a cover.</p>
<p>The consequence is that every later extension of the model inherits the same blindness, since the zones and conduits of IEC 62443 describe communication paths between assets, Zero Trust verifies flows and microsegmentation ties policies to equipment identities, so none of these mechanisms has any hold over a person standing in front of an open cabinet, as we <a href="/blog/with-all-purdue-respect">showed in the second article</a>.</p>
<h2 id="this-is-not-an-edge-case">This is not an edge case</h2>
<p>One might object that physical intervention is the exception and logical access the rule, but from what we observe it is rather the reverse, because on most industrial sites physical intervention is the predominant path for modifying a level 0 or level 1 device, since a large share of the PLCs, drives, controllers and test benches in service today cannot be configured any other way and you have to be standing in front of them, with the right cable and the right software, and at best a logbook is kept on site, or else one hopes to remember to record the day&#39;s intervention in the IT tool once back at the office, with all the approximations that entails.</p>
<p>These are, moreover, often the very devices that the operator has decided, often rightly, will never be connected to the IT network, which is what is called an air gap, precisely because the connection would create more risk than it would address, and the result is paradoxical, because the more isolated a device is from the network, the better protected it is against network attacks and the less it is governed at all, so isolation does not remove the need for access control but only the usual means of exercising it.</p>
<p>Nor are the causes for concern the ones people imagine, because a large share of the people legitimately standing in front of a machine do not belong to the organisation, whether temps, tier-two subcontractors, manufacturer&#39;s technicians who have never been on site before or service providers engaged by the maintainer itself, and they appear in no directory, while there is often no directory on the supplier&#39;s side to federate with, as we wrote about external contractors.</p>
<h2 id="what-the-standards-already-acknowledge-about-physical-access">What the standards already acknowledge about physical access control</h2>
<p>The most interesting point is that the body of standards says all of this, sometimes with disarming candour.</p>
<p>IEC 62443-3-2 lays down the rule that makes the subject unavoidable, because all communication between zones passes through an explicitly defined and monitored conduit, so if a connection exists it is a conduit and it is controlled, and a technician with a laptop and a programming cable is a connection between the outside world and a PLC, which nobody treats as such although it meets, in our view, the definition.</p>
<p>IEC 62443-4-2 goes further in its definition of a compensating countermeasure, since it gives the example of physical access control to a control room, intended to restrict access to a group of known personnel, and states its objective, which is to compensate for the technical requirement for the industrial system itself to uniquely identify people, so the standards committee writes that physical access control is what gets deployed when the industrial system is incapable of telling one human being from another, which is to say more or less always.</p>
<p>NIST SP 800-82 Revision 3 places physical security in its layered defence-in-depth model, and its appendix on tailoring controls to operational technology is unambiguous about control PE-3, because physical access controls complement OT security when electronic mechanisms are unable to meet security requirements, so two major frameworks, both compatible with Purdue, say the same thing, namely that when logical identification is impossible the physical takes over, and neither of them says how.</p>
<p>So the subject is not absent from the standards, since IEC 62443-3-3 includes a system requirement devoted to controlling the use of portable and mobile devices and IEC 62443-2-4 governs the practices of service providers, both of them squarely addressing the scene we have just described, and we have detailed elsewhere what the standard asks. These requirements describe what ought to be achieved, controlled use, verified authorisation and a trace, without saying by what means to achieve it on a device that has no screen, no identity and no network, and in practice they are met through procedures, a reception instruction, a paper register or a clause in the maintenance contract, but a procedure refuses nothing, measures nothing and signs nothing, so the result is a requirement that is formally covered and materially unverifiable.</p>
<h2 id="the-purdue-missing-link-and-where-trust-buster-stands">The Purdue missing link, and where Trust-Buster stands</h2>
<p>Here is our answer to the opening question, and it is somewhat uncomfortable to put into words, because we are at no level of the model but upstream of level 0.</p>
<p>Level 0 is the physical process, the motors, valves, sensors and actuators, but there is a precondition for the very existence of that level, which the model does not draw because it went without saying in 1992, and that is energy, because a machine without power does not produce, does not heat, does not turn, does not communicate and cannot be reprogrammed.</p>
<p>A policy enforcement point placed on the power supply therefore has a property that no software enforcement point can have, because it applies before the target exists functionally, while a software enforcement point runs after the target has started up and depends on its integrity, and no vulnerability in the device&#39;s firmware makes it possible to get around the absence of power, as long as the power really does go through that point and that point alone.</p>
<p>This is also what makes the approach agnostic, because it makes no assumption about the protocol, the firmware, the age, the manufacturer or the presence of a network stack, so the control applies in the same way to a PLC from 1998 and to a machine delivered last month.</p>
<p>If you prefer the vocabulary of zones and conduits to that of levels, the wording is sharper, because we draw the human conduit, this Purdue missing link, and we give it a rule, an enforcement point and a log, whereas it already existed and was governed by nothing.</p>
<h2 id="the-mapping-level-by-level">The mapping, level by level</h2>
<p>For those who have to present this to an architecture board.</p>
<table>
<thead>
<tr>
<th>Purdue level</th>
<th>What usually protects it</th>
<th>What we bring to it</th>
</tr>
</thead>
<tbody><tr>
<td>5, enterprise</td>
<td>IAM, firewalls, endpoint detection</td>
<td>Nothing, this is not our field.</td>
</tr>
<tr>
<td>4, site business planning</td>
<td>Same</td>
<td>Nothing.</td>
</tr>
<tr>
<td>3.5, industrial DMZ</td>
<td>Protocol break, data diode, bastion host</td>
<td>Nothing, it is probably the most consequential control in the architecture and it is held by others.</td>
</tr>
<tr>
<td>3, site operations</td>
<td>Segmentation, monitoring</td>
<td>Nothing.</td>
</tr>
<tr>
<td>2, area supervisory control</td>
<td>HMI hardening, logging</td>
<td>Physical intervention on a supervisory workstation can be logged.</td>
</tr>
<tr>
<td>1, basic control</td>
<td>Inline isolation, flow control</td>
<td>The human conduit to the PLC is drawn and governed.</td>
</tr>
<tr>
<td>0, physical process</td>
<td>Nothing, in practice</td>
<td>The enforcement point is upstream, on the power supply.</td>
</tr>
</tbody></table>
<h2 id="what-we-do-not-do">What we do not do</h2>
<p>A reader who has just gone through these articles on Purdue will legitimately wonder whether we claim to replace their architecture, and the answer is no.</p>
<p>We do no network segmentation, so no industrial firewall, no microsegmentation and no east-west policy, and we do not hold the boundary between levels 3 and 4, because the demilitarised zone, the protocol break and unidirectional gateways are other trades, practised by serious players, and that boundary very probably remains the most consequential control in your architecture.</p>
<p>Nor do we provide a safety function, because cutting a power supply is, for us, a security control and not a certified function, and it is no substitute for an emergency stop, a lockout or a safety relay, as we have written in detail.</p>
<p>What we do fits in one sentence, because it is about controlling and proving who energises a piece of equipment, across an estate that has no screen, no identity and no network, including when nothing else responds, and we intend to offer other means of control at level 0 with serial, USB or BLE effectors, so as to be able to detect, identify, verify and authorise any intervention by a technician on the estate, but the only fully operational effector to date is the power cut.</p>
<h2 id="the-model-is-not-missing-a-level">The model is not missing a level</h2>
<p>The temptation would be to propose an additional level, a &quot;level minus one&quot; for the human and for energy, and it would be elegant but probably wrong, because the Purdue missing link is not a tier in the hierarchy, it runs through it, since the same technician who opens a level 1 cabinet may, in the afternoon, log on to a level 2 supervisory workstation and plug a USB stick into a level 3 server, so it is an orthogonal axis and not a layer.</p>
<p>So the point is not so much to complete the diagram as to acknowledge that an architecture which describes only machines cannot, on its own, govern what people do.</p>
<p><em>Last article in the series: <a href="/blog/remote-access-and-physical-access-two-different-problems">the same technician, with the same laptop</a> and the same ability to reprogram, sometimes connects remotely from another continent, and that case is well handled, but by a different trade from ours, so we mention it to make the distinction quite explicit.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>With All Purdue Respect: Three Breaches and a Blind Spot</title>
      <link>https://trust-buster.com/blog/with-all-purdue-respect</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/with-all-purdue-respect</guid>
      <pubDate>Fri, 11 Sep 2026 10:02:35 GMT</pubDate>
      
      <description>IIoT, cloud and remote access bypass the Purdue model, the trade answered with network controls, and one path stays in the Purdue blind spot, the person.</description>
      <content:encoded><![CDATA[<p>The Purdue model has been called into question since the early 2020s for three sound reasons, because the industrial internet of things, or IIoT, cloud connectivity and remote access all bypass its hierarchy, and the profession has answered with solid controls, but one path remains in what we will call here the Purdue blind spot, because it never appeared on the diagram and nobody has corrected it.</p>
<p>The answers usually put forward, that is IEC 62443 zones and conduits, Zero Trust, microsegmentation and unidirectional gateways, are solid and deserve to be known, but they all share one feature that is rarely noticed, since each describes a network path the model had not anticipated and each responds to it with a network control.</p>
<p><em>This article does not claim that the Purdue model is dead, and it only tries to look at what the profession has corrected in recent years, and at what nobody has corrected, which is where Trust-Buster can improve the situation.</em></p>
<h2 id="the-three-breaches-without-indulgence">The three breaches, without indulgence</h2>
<h3 id="the-iiot-sensor-that-talks-to-the-cloud">The IIoT sensor that talks to the cloud</h3>
<p>The model rests on a simple assumption, which is that a level only talks to its neighbours, but a connected sensor that sends its measurements straight to an online service bypasses four levels at once, and this is not a textbook case, because a compressor delivered with its built-in telemetry board, a machine whose manufacturer provides predictive maintenance from its own servers or a vibration sensor installed by a contractor with its own cellular link all arrive on site bringing their own way out, and they do not ask the hierarchy for permission since they ignore it.</p>
<h3 id="remote-access-now-the-norm">Remote access, now the norm</h3>
<p>Remote working, remote maintenance and the growing scarcity of skills have made remote access an operational necessity rather than a convenience, because a German manufacturer no longer travels for an adjustment, yet the original model had no place for it, so it was grafted on, often in a hurry and sometimes with a general-purpose remote control tool installed on a supervision workstation.</p>
<p>The consequences are not a matter of conjecture, since on 22 January 2026 CERT-FR published <a href="https://www.cert.ssi.gouv.fr/dur/CERTFR-2025-DUR-003/">a recommendation</a> stating that ANSSI had observed several compromises of French entities in the renewable energy generation and water management sectors, and hacktivist groups achieved them by remotely accessing management interfaces of industrial installations exposed on the internet, the most successful action having shut down a wind farm for a few hours with a financial loss for the operator. The agency specifies that these activities are most often of a low technical level and that the equipment concerned was exposed without authentication or with a default password, some industrial protocols being directly reachable from the internet without encryption, so this is a path the model had not anticipated, exploited without any particular skill, on real and recent installations.</p>
<h3 id="lateral-movement-once-inside">Lateral movement, once inside</h3>
<p>The hierarchy protects the boundaries between levels, but within a zone it protects nothing, and an attacker who has gained a foothold on a level 2 workstation moves freely towards the other systems on the same level, which is called lateral movement, because the model never said what should happen there, and this is a structural flaw rather than an oversight, since Purdue describes the vertical flow of data and not the horizontal one.</p>
<h2 id="the-profession-s-answers-and-they-are-good-ones">The profession&#39;s answers, and they are good ones</h2>
<p>Industrial security deserves credit for not standing idle, because four answers have established themselves within a few years and they work.</p>
<p>IEC 62443-3-2 zones and conduits extend the model rather than replace it, because a zone groups assets by security requirement rather than by level, so a safety instrumented system and a supervision interface can sit at the same Purdue level while belonging to different zones, since the consequences of compromising them have nothing in common, and each zone is given a target security level, while every communication path between zones becomes a conduit that must be identified, justified and secured. The contribution seems important to us, because this mechanism makes it possible to bring into the architecture the modern flows that the hierarchy could not describe, and replication to the cloud, remote access or edge messaging become conduits as soon as a connection exists, so they are controlled.</p>
<p>Zero Trust modernises these conduits, since rather than a single perimeter, every flow is verified at every connection, and software microsegmentation can replace hand-configured VLANs, with policies based on cryptographic device identities rather than physical ports.</p>
<p>The boundary between levels 3 and 4 is hardening, and this is the most widely agreed point in the field, because the industrial demilitarised zone probably remains the most consequential control in an OT architecture and it is treated seriously, with a genuine protocol break, hardware unidirectional gateways and outbound-only connectivity.</p>
<p>And for equipment too old to carry an agent, inline isolation offers an elegant answer, because a hardware box placed in series on the network cable in front of the PLC inspects traffic and enforces the conduit&#39;s rules without installing anything on the legacy system and without re-architecting the network.</p>
<p>We have nothing to object to in any of this, because it is good work on real problems.</p>
<h2 id="the-purdue-blind-spot">The Purdue blind spot</h2>
<p>Now go back over the list and look at what these elements have in common, because the sensor that talks to the cloud is a network path, remote access is a network path, lateral movement is a network path, and microsegmentation, the bastion host, that is the jump server everyone is forced through, the unidirectional gateway and inline isolation are network controls.</p>
<p>Every critique levelled at the Purdue model in recent years therefore describes a communication path it had not anticipated, and every answer offered is a control placed on a communication path, and this is not a reproach, because it is consistent, since the model describes data flows and it is therefore natural that it should be criticised for missing data flows and that the response should be to control data flows.</p>
<p>One player in this market, incidentally, states the limitation candidly, because in its analysis of the Purdue model and Zero Trust, Brixio writes that this approach modernises the conduits and the IT-OT boundary, but that it does not change the physical topology of levels 0 to 2 and that &quot;it does not touch the PLCs&quot;, and that is precisely the Purdue blind spot, because the person who touches the PLC uses none of these paths.</p>
<h2 id="the-model-is-not-outdated-it-is-incomplete">The model is not outdated, it is incomplete</h2>
<p>The usual conclusion of articles on the subject is that Purdue is obsolete, but we do not think so, because the authoritative frameworks continue to rely on it, since IEC 62443 anchors its zones and conduits in it, NIST SP 800-82 revision 3 uses it to describe its defence in depth and the reference architectures of the major vendors derive directly from it, as we <a href="/blog/purdue-diligence">recalled in the first article</a>, and what it is really criticised for is failing to describe paths that did not exist when it was written.</p>
<p>But there remains one, older than all the others, which has never appeared on the diagram and which nobody talks about, because it does not go through a port and does not carry packets, and on most industrial sites it is still the one used most often.</p>
<p>The US doctrine of November 2025 illustrates this ambivalent position well, because it asserts that the Purdue model, IEC 62443 and the UFC 4-010-06 criteria remain authoritative frameworks for classifying OT systems, and in the same document it chooses not to use them, preferring an abstract two-layer topology judged more flexible because it does not assign components and users to rigid levels, so the vocabulary still carries authority while the grid is starting to get in the way, and the Purdue blind spot in both of them is still the person.</p>
<p><em>Next in the series: <a href="/blog/long-purdue">a technician who opens a cabinet</a> and plugs a laptop into a PLC&#39;s programming port has not crossed a single firewall, yet they have entered level 1 directly, and that is the subject of the third article.</em></p>
<p><em>Sources:</em></p>
<ul>
<li>Department of War, Chief Information Officer, Zero Trust Portfolio Management Office. <a href="https://dodcio.defense.gov/Portals/0/Documents/Library/ZT-OperationalTechnologyActivitiesOutcomes.pdf"><em>Zero Trust for Operational Technology Activities and Outcomes</em></a>. 18 November 2025.</li>
<li>Brixio. <a href="https://brixio.io/blog/purdue-model-zero-trust-ot/">&quot;Purdue model &amp; Zero Trust: modernising OT segmentation&quot;</a>. June 2026.</li>
<li>ANSSI / CERT-FR. <a href="https://www.cert.ssi.gouv.fr/dur/CERTFR-2025-DUR-003/"><em>CERTFR-2025-DUR-003, Recommendations for stakeholders in the energy and water sectors</em></a>. 22 January 2026.</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>Is Radio Monitoring in a Workplace Legal?</title>
      <link>https://trust-buster.com/blog/is-radio-monitoring-in-a-workplace-legal</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/is-radio-monitoring-in-a-workplace-legal</guid>
      <pubDate>Thu, 10 Sep 2026 16:08:59 GMT</pubDate>
      
      <description>Is radio monitoring in a workplace legal? Europe treats a hardware address as personal data, the rules differ by country, and here is how we design for it.</description>
      <content:encoded><![CDATA[<p>Radio monitoring in a workplace means processing personal data, because a device&#39;s hardware address, the MAC address that every Bluetooth or Wi-Fi chip carries, has been treated as personal data by European supervisory authorities for over a decade, and what follows from that differs more from one country to the next than the marketing in this field lets on, so we say up front that we are engineers and not lawyers, and that what follows is how we understand the rules we design against rather than legal advice.</p>
<p><em>Last of four articles on radio detection in industrial environments, after <a href="/blog/who-is-in-the-room">what you can hear</a>, <a href="/blog/what-an-sdr-finds">what you can find</a> and <a href="/blog/jamming-relays-and-the-flipper-zero">how to tell when you are being deceived</a>, and this one covers whether you are allowed to do any of it.</em></p>
<h2 id="why-this-article-exists">Why this article exists</h2>
<p>Almost every technical article on radio detection skips this chapter, and we think that is a mistake with a predictable ending, because a system deployed on sound engineering and then discovered by a workforce that was never told tends to be removed within a quarter. We prefer to treat the question as an engineering constraint rather than a legal afterthought, because the rules shape the architecture and not only the paperwork.</p>
<h2 id="the-european-position-and-the-cnil-guidance">The European position and the CNIL guidance</h2>
<p>A device&#39;s hardware address is personal data in the eyes of European supervisory authorities, who have been consistent on this point since well before the GDPR, and the reasoning is simple enough, because an identifier that persists from one observation to the next and attaches to a device carried by one person singles out that person whether or not you know their name. The most detailed published guidance we know of in Europe comes from the French supervisory authority, the CNIL, and it is worth knowing wherever you operate because it remains, as far as we can tell, the most specific document on this exact use case, and in substance it asks the following. For simple aggregate counting it asks for anonymisation within five minutes of collection, and the word is anonymisation rather than pseudonymisation, which means the operation has to be irreversible, or alternatively for immediate pseudonymisation followed by anonymisation or destruction within twenty-four hours, together with a markedly fuller notice to the people concerned. It asks for visible signage on site, on the model of CCTV notices, pointing to fuller information, and it asks for a workable right to object, and on this last point one clarification settles much debate, because asking people to switch off their phone&#39;s Wi-Fi was not considered an acceptable way to object, since it deprives them of a function of their own device. The history is instructive too, because in 2015 a large outdoor advertising company was refused authorisation for a Wi-Fi counting system in a Paris business district on the grounds that the anonymisation it proposed was reversible, as the operator could replay the hashing process, and reversible anonymisation is pseudonymisation under a better name, and the regulator was not taken in by it.</p>
<h2 id="where-the-picture-differs">Where the picture differs</h2>
<p>Two caveats are worth stating, because the field is not uniform and vendors who sell globally tend to imply that it is. Outside the EU and the UK the position is far less settled as far as we know, because there is no US federal equivalent to the GDPR covering this case, some state privacy laws include device identifiers in their definition of personal information while others do not, and sector rules apply where they apply and stay silent elsewhere. That looks like a compliance advantage today and a liability tomorrow, because the privacy regimes we have watched have tightened rather than loosened, so a system built to exploit a permissive jurisdiction will probably need rebuilding when that jurisdiction catches up. The workforce dimension varies substantially as well, because in several European jurisdictions deploying radio monitoring in a workplace triggers a formal employee consultation obligation before installation and not after, while in others there is no such requirement, and our recommendation is the same either way, which is to treat it as an obligation, because a system discovered by the workforce rather than presented to it rarely survives its first meeting with staff representatives, and the engineering merits of the architecture will not be part of that conversation.</p>
<h2 id="what-we-do-about-radio-monitoring-in-a-workplace">What we do about radio monitoring in a workplace</h2>
<p>Our position follows from the architecture more than from virtue, and we think that is the only kind of privacy claim worth making. We do not need to track strangers, because the system answers one question, which is whether the enrolled holder is present, so the identifiers we retain are those of voluntarily enrolled badges and phones whose holders are informed and whose pairing key was handed to us explicitly, and for everything else, including the anomaly detection described in the third article, an aggregate counter and a signature comparison are enough. The rotating address mechanism, which phone makers call MAC randomization, works in our favour, because as we explained in <a href="/blog/who-is-in-the-room">Who Is in the Room?</a> an unknown Bluetooth address cannot be linked from one cycle to the next without the pairing key, so not tracking third parties is a property of the protocol rather than a contractual promise, and when someone asks whether our system tracks visitors&#39; phones the answer is that it cannot, which we find better than a promise that it does not. Our logs name authorised people and not passers-by, because an anomaly event records that an unidentified emitter was present, with its timestamp and its band, and it does not reconstruct a third party&#39;s identity since it has no means to. The same rule applies to our keyboard and mouse effector, which logs the event, whether a connection, a cut or an unauthorised pairing attempt, and never the content, because recording keystrokes builds a keylogger however good the intention, and we doubt any security justification would survive that description in front of a data protection officer.</p>
<h2 id="three-questions-to-ask-any-vendor-in-this-field">Three questions to ask any vendor in this field</h2>
<p>They are useful whoever you are evaluating, including us. The first is what identifiers you retain and for how long, because if the answer involves storing raw hardware addresses of unenrolled devices for any period beyond immediate processing you should ask what the lawful basis is and expect a precise answer. The second is whether the system can track a device you have not enrolled, because if it can that is a capability that has to be justified and governed and disclosed, while if it cannot as a matter of protocol rather than policy that is worth considerably more. The third is what appears in the log when an unidentified device is detected, because a timestamp, a band and a signature class is the kind of answer we would expect, while anything from which a person could be re-identified is a wrong one. If a vendor cannot answer these three without checking with someone, the system was probably designed before the question was asked.</p>
<h2 id="the-engineering-point">The engineering point</h2>
<p>The thread running through this series is that radio detection corroborates a decision rather than anchoring one, and the data protection position follows from exactly the same place, because a system that needs to identify everyone in a room in order to work has taken on a legal burden it did not need, while a system that identifies only those who volunteered to be identified and treats everything else as an unattributed anomaly signal does the same operational job with a fraction of the exposure. That is not a compromise we made under regulatory pressure, since it is the architecture that fell out of asking what the system needed to know, which turned out to be much less than what it could know.</p>
]]></content:encoded>
    </item>
    <item>
      <title>Jamming, Relays and the Flipper Zero</title>
      <link>https://trust-buster.com/blog/jamming-relays-and-the-flipper-zero</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/jamming-relays-and-the-flipper-zero</guid>
      <pubDate>Thu, 10 Sep 2026 16:04:39 GMT</pubDate>
      
      <description>A jammed band and a broken receiver look the same from inside, so we refuse rather than decide blind, and we describe the Flipper Zero as it is.</description>
      <content:encoded><![CDATA[<p>A jammed radio band and a broken receiver look identical from the inside, and telling them apart is the hardest problem in radio-based access control, which is why our system treats silence as a refusal rather than as a permission, and it is also why the Flipper Zero, which comes up in every conversation on this subject, deserves to be described precisely rather than feared.</p>
<p><em>Third of four articles on radio detection in industrial environments, after <a href="/blog/who-is-in-the-room">Who Is in the Room?</a> and <a href="/blog/what-an-sdr-finds">What an SDR Finds</a>, which covered what you can hear, and this one covers what happens when someone does not want you to hear it, or wants you to hear the wrong thing.</em></p>
<h2 id="three-families-of-jamming">Three families of jamming</h2>
<p>The first family is wideband noise, where someone transmits power across the whole band continuously, and it is simple and effective and perfectly visible to anyone who looks. The second is reactive jamming, where the transmitter listens and only emits when it detects the start of a legitimate frame, just long enough to corrupt it, and it is far more discreet because the noise floor stays entirely normal. The third is protocol jamming, where no noise is emitted at all but only valid frames, such as the forced Wi-Fi disconnections known as deauth frames, so nothing is abnormal at the radio level while everything is abnormal at the logical level. Any detection strategy that only looks for the first family will miss the other two, and in our experience that describes most of what gets deployed.</p>
<h2 id="the-giveaways">The giveaways</h2>
<p>None of these signs is decisive on its own, and it is their conjunction that makes a signal. A noise floor that rises with no decodable traffic is the classic signature, because plenty of energy is received and no valid packet comes out of it, so something is transmitting that does not speak our language. An integrity-check failure rate that climbs sharply is another, because packets arrive but their checksums fail, which means they are being corrupted in flight, and this is the signature of reactive jamming that the noise floor does not show. Channel occupancy that departs from its baseline is a third, because every room has its own occupancy signature, learned during the observation phase before anything is armed, and a marked departure from that reference says more than any absolute threshold, since thresholds that work in an electronics lab mean nothing in a machine shop. The most useful, in our view, is simultaneous disappearance, because our effectors, the small units placed beside the machines, transmit periodically, and if they all stop being heard within the same second, four hardware failures at once is markedly less likely than something covering the band. It costs nothing, since it falls out of the system talking to itself continuously, with no additional sensor or processing, so the most reliable jamming signal we have is a by-product of the architecture rather than a feature we designed.</p>
<h2 id="what-we-do-about-it-and-what-it-costs">What we do about it, and what it costs</h2>
<p>Nothing more than a refusal, and deliberately so, because you cannot prevent jamming but you can refuse to decide blind, so detected jamming drops the zone to refusal with a logged event that carries its reason, and nothing in the system replies on the air, since we never transmit against what we hear. The trade-off has to be stated plainly to the operator before the contract rather than after, because the system is susceptible to denial of service, and someone who wants to stop your production rather than enter it will manage by covering the band. The answer to that is organisational rather than technical, with a derogation procedure, a fallback mode and monitoring of radio availability, because at that point we are up against the equivalent of a pair of bolt cutters, and that is resolved by building access control and sound personnel management. It is also a matter for the law, because as far as we know possessing and operating radio jammers is prohibited in most jurisdictions, including the EU, the United Kingdom and the United States where the FCC treats it as a federal offence, which does not deter everyone but changes the nature of the incident from a security failure into a criminal act. A closing note on direction, because an access control that opens on failure has become a green light with nothing behind it, so the direction of failure was chosen at design time rather than suffered, and it is the first property we verify in any deployment.</p>
<h2 id="being-deceived-rather-than-blinded-from-replay-to-relay-atta">Being deceived rather than blinded, from replay to relay attack</h2>
<p>Jamming removes information, while spoofing supplies false information, and it is the harder problem. The replay attack is the part that is solved, because a BLE advertisement travels in the clear and copies easily, which is why we never treat an advertisement as proof of identity, since it opens a conversation and does not authorise one, and authorisation comes from a cryptographic challenge that only the key inside the badge&#39;s secure element can answer and that cannot be replayed because it changes every time. That parry is reliable and well understood, and it is also the last easy answer in this article. The relay attack is the part that is not solved, because an attacker who places one device near a legitimate badge left in the locker room and a second near the machine, and lets the conversation pass between them, forges nothing at all, since the messages are authentic and the signatures valid and the challenge correctly solved, and what is manufactured is proximity, and proximity is not a property you can sign. No cryptographic layer sees this attack and only a physical measurement contradicts it, and time-of-flight ranging over ultra-wideband, or UWB, is the benchmark, because light does not accelerate on request so a relay necessarily adds delay, and it is the only structural answer we know of, and we do not have it today and will not pretend otherwise, since it is a roadmap line and our most visible gap against the Aliro standard published by the Connectivity Standards Alliance. Meanwhile two things raise the cost without closing the gap. The first is multi-sensor coherence, because with several receivers spread through a room a real emitter produces a level profile consistent with its position, where a relay often produces an inconsistency, such as a very strong signal seen by one sensor only or a profile that does not change while the person moves, and that is not a proof but an implausibility, and an implausibility is enough to refuse. The second is the surrounding conditions, because presence must be maintained throughout the period of use, the holder&#39;s time window must be open at the same moment, and the access is logged under their name. There is finally the emitter&#39;s radio fingerprint, because every transmitter carries measurable manufacturing imperfections, such as frequency offsets, phase noise and modulation defects, which form a signature that is hard to imitate, an active and promising research field that we follow and do not sell.</p>
<h2 id="the-flipper-zero-precisely">The Flipper Zero, precisely</h2>
<p>Since it comes up so often, and usually wrongly, here are the facts as we understand them. What it does is carry a sub-GHz transceiver covering the 300 to 348, 387 to 464 and 779 to 928 MHz bands, with an advertised range of about fifty metres, and on those bands it reads, records and replays fixed-code remotes, such as gates, barriers, radio switches and doorbells, so against fixed-code equipment it works well, and it has made common a skill that previously required real expertise. What it does not do is touch 2.4 GHz, so no Wi-Fi, no Bluetooth and no Zigbee, and its internal power stage is weak enough that a jamming attempt carries only a few tens of centimetres, so the Flipper Zero alone is not the universal threat it is often described as. What changes the picture is the add-on boards sold for a few tens of euros, which add a Wi-Fi module, a proprietary 2.4 GHz module and an amplified sub-GHz stage with an external antenna, and with that set the device reaches wireless keyboard and mouse attacks, Wi-Fi attacks and much longer-range sub-GHz jamming, so the threat worth planning for is less the Flipper Zero at a hundred and seventy euros than the Flipper Zero with its thirty-euro board. How it is seen is what matters to us, because each module leaves a trace, since the sub-GHz emission is visible to an SDR effector, the most common Wi-Fi firmware on these boards produces deauthentication frames in a characteristic volume, and the proprietary 2.4 GHz module transmits on distinctive channels with a distinctive addressing format. None of these is proof of intent, but together, in a workshop where nothing of the sort should be, they form an anomaly worth an alert and nothing more, because a detection system that concludes attack from a radio signature will eventually accuse an innocent contractor&#39;s tooling and be switched off soon after.</p>
<h2 id="what-detection-proves">What detection proves</h2>
<table>
<thead>
<tr>
<th>Mechanism</th>
<th>Proven</th>
<th>Not proven</th>
</tr>
</thead>
<tbody><tr>
<td>Rising noise floor, no valid traffic</td>
<td>That the band became unusable</td>
<td>Who is jamming, or from where</td>
</tr>
<tr>
<td>Integrity failure rate</td>
<td>Frames are being corrupted in flight</td>
<td>Whether deliberately</td>
</tr>
<tr>
<td>Simultaneous effector silence</td>
<td>Something is covering the band</td>
<td>That it is an attack rather than an environmental event</td>
</tr>
<tr>
<td>Multi-sensor coherence</td>
<td>A geometric implausibility</td>
<td>A relay with certainty, which only time of flight gives</td>
</tr>
<tr>
<td>Flipper module signatures</td>
<td>A device of this class is transmitting nearby</td>
<td>Intent</td>
</tr>
</tbody></table>
<p>Every row in the right-hand column is a reason to raise an event and refuse, and never a reason to accuse, because radio detection corroborates a decision and never anchors one.</p>
<h2 id="hearing-is-not-understanding">Hearing is not understanding</h2>
<p>An advertisement received does not prove a presence, and a presence does not prove an identity, and an identity does not prove an authorisation, and silence does not prove an absence either, because it may also mean that someone put a cushion over our ears. That is why our system treats silence as a refusal rather than as a permission, and it is the least spectacular decision in the whole architecture, and the one we would defend first.</p>
<p><em>Last in the series, <a href="/blog/is-radio-monitoring-in-a-workplace-legal">Is Radio Monitoring in a Workplace Legal?</a>, because everything described across these three articles involves listening to radio traffic in a workplace, and that is not legally neutral.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>What an SDR Finds</title>
      <link>https://trust-buster.com/blog/what-an-sdr-finds</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/what-an-sdr-finds</guid>
      <pubDate>Thu, 10 Sep 2026 16:01:53 GMT</pubDate>
      
      <description>An SDR sweeps and misses what transmits while it looks elsewhere, so here is what wideband listening finds in a workshop and what it cannot promise.</description>
      <content:encoded><![CDATA[<p>A single-channel software-defined radio, an SDR, cannot listen to a wide band all at once, because it sweeps, pausing a few milliseconds on one portion of the spectrum before moving on, and so it misses everything that transmits while it looks elsewhere, and that is a structural limit rather than a tuning fault, and it is the first thing to understand before buying anything.</p>
<p><em>Second of four articles on radio detection in industrial environments, after <a href="/blog/who-is-in-the-room">Who Is in the Room?</a>, which covered Bluetooth and Wi-Fi presence detection, and this one goes beyond them.</em></p>
<h2 id="what-a-software-defined-radio-or-sdr-is">What a software-defined radio, or SDR, is</h2>
<p>An SDR is a receiver that does nothing on its own, because it digitises a raw slice of spectrum and hands it to a computer, which decides how to interpret it, so the same hardware listens to a gate remote, a temperature sensor, a contactless badge or a cellular modem depending only on the program you give it. That is what makes a dedicated listening effector realistic, since it comes down to an SDR card on a single-board computer, with the antenna suited to the target band, running software that decodes what matters and discards the rest, and we only ever use it to receive.</p>
<h2 id="choosing-hardware-and-a-trap-we-fell-into">Choosing hardware, and a trap we fell into</h2>
<table>
<thead>
<tr>
<th>Hardware</th>
<th>Coverage</th>
<th>Note</th>
</tr>
</thead>
<tbody><tr>
<td>RTL-SDR (about 30 euros)</td>
<td>about 24 MHz to 1.76 GHz</td>
<td>Best value by a wide margin, but it does not reach 2.4 GHz, so it sees neither Wi-Fi nor Bluetooth</td>
</tr>
<tr>
<td>HackRF One</td>
<td>1 MHz to 6 GHz</td>
<td>Covers everything, including 2.4 and 5 GHz, with an 8-bit converter that makes it less sensitive</td>
</tr>
<tr>
<td>Airspy R2</td>
<td>24 to 1800 MHz</td>
<td>Better reception than the RTL-SDR on the same band</td>
</tr>
<tr>
<td>LimeSDR / USRP</td>
<td>up to 3.5 to 6 GHz</td>
<td>Considerably more expensive, laboratory quality</td>
</tr>
<tr>
<td>KrakenSDR</td>
<td>24 to 1766 MHz</td>
<td>Five receivers synchronised on one clock, for direction finding</td>
</tr>
</tbody></table>
<p>The trap is in the first row, and we fell into it like everyone else, because many projects start with an RTL-SDR expecting to monitor all radio and then discover that the most interesting band in an industrial setting, the 2.4 GHz where Wi-Fi, Bluetooth and Zigbee live, is precisely out of its reach. Serious wideband listening therefore needs at least two radios, one for sub-GHz and one for 2.4 GHz, and any product or project claiming full coverage on a single thirty-euro receiver has, in our view, not looked closely at its own specification.</p>
<h2 id="what-you-actually-find">What you actually find</h2>
<p>On the sub-GHz bands the reference tool is rtl_433, a free decoder that recognises over a thousand commercial sensor and remote protocols across 315, 433, 868 and 915 MHz, and what it reveals in a workshop is usually a surprise, and rarely a pleasant one, because you find the following.</p>
<ul>
<li>Temperature and humidity probes fitted by a contractor during a survey and never removed</li>
<li>Crane and gate remotes, often on fixed codes</li>
<li>Door and window sensors from a previous alarm system</li>
<li>Weather stations, and tyre pressure sensors from vehicles in the yard</li>
<li>And, with some regularity, a piece of equipment that appears on no inventory at all</li>
</ul>
<p>That last category is the interesting one, because an estate whose emitters are unknown is an estate whose asset inventory is incomplete, and asset inventory is an explicit requirement under NIS2 and an expected control under most critical-infrastructure regimes we know of, so this is where radio detection stops being a security curiosity and becomes a compliance instrument, since a wideband sweep is, among other things, the fastest way we know to discover what a decade of contractors left behind.</p>
<h2 id="the-limit-sweeping-is-not-listening">The limit: sweeping is not listening</h2>
<p>A single-channel SDR sees one window of spectrum at a time, typically a few megahertz wide, and to cover a broad band it sweeps, so the consequence is direct, because it misses everything transmitting while it looks elsewhere, and a remote that transmits for three hundred milliseconds during a two-second sweep cycle stands a good chance of going entirely unseen, and if you repeat the sweep you may catch it next time or you may not, if it only ever transmits once. This cannot be tuned away, because narrowing the swept range improves the odds and reduces the coverage while widening it does the reverse, and there is no setting that gives both. Professional systems get round it with numbers, which means several receivers in parallel, each dedicated to a band and all listening continuously, and Bastille for instance advertises continuous coverage from 25 MHz to 6 GHz and permanent demodulation of every Bluetooth channel where an ordinary sniffer follows only a few, which is a difference in kind rather than in degree and explains the price gap between a thirty-euro dongle and a deployed sensor network. We therefore do not claim exhaustive wideband detection, because an SDR effector serves to inventory and to flag anomalies and not to guarantee that no emitter escaped us, and that guarantee does not exist as far as we know, here or anywhere, so a vendor offering it is describing an ambition rather than a measurement.</p>
<h2 id="direction-finding-and-why-it-matters-more-than-it-sounds">Direction finding, and why it matters more than it sounds</h2>
<p>Knowing that something transmitted is useful, and knowing where it transmitted from is considerably more useful, and until recently it was expensive. The KrakenSDR changed that, with five RTL-SDR receivers sharing a single clock and integrated calibration hardware, processed by correlative interferometry, with open-source acquisition and processing software running on a Raspberry Pi, and its limit for our purposes is blunt, because it reaches 1.766 GHz and therefore does not cover Bluetooth&#39;s 2.4 GHz, so for direction finding in BLE the standardised path is Bluetooth 5.1 angle of arrival using an antenna array, which is a different piece of hardware entirely. Direction matters because an emitter that appears at two incompatible locations, or whose apparent bearing does not change while the person supposedly carrying it walks across a room, is producing an inconsistency, and that is not a proof of attack but an implausibility, and in an access control decision an implausibility is enough to refuse.</p>
<h2 id="where-this-sits-in-the-market">Where this sits in the market</h2>
<p>The radio detection market reads on three floors, and it is honest to say which one is ours. At the top sits spatial surveillance, where Bastille Networks is the reference, with continuous coverage across a very wide band, networks of passive PoE sensors, emitter location to within a few metres and movement history over time, and it is an installation-security product sold to government and large enterprise and priced accordingly. In the middle sits Wi-Fi intrusion detection built into the major vendors&#39; network infrastructure, effective within its perimeter and blind to everything else, whether sub-GHz, proprietary protocols or anything not speaking 802.11. At the bottom sits free tooling, such as rtl_433, Kismet, OpenMQTTGateway, the Nordic sniffer and KrakenSDR, which is amply sufficient to build an honest sensor if you know what you are doing and what you are not measuring. We are not on the first floor and do not claim to be, because our object is not to watch a space but to corroborate an access decision, locally, without a network, beside a machine, and a spatial surveillance system tells you a phone entered the building while it does not cut power to the press, so these are two trades, and they coexist happily on the same site.</p>
<h2 id="a-small-toolbox">A small toolbox</h2>
<ul>
<li>Sub-GHz: <a href="https://github.com/merbanan/rtl_433">rtl_433</a> and its decoders, on RTL-SDR or Airspy</li>
<li>Wideband: HackRF One to reach 2.4 and 5 GHz, with GNU Radio for processing</li>
<li>Direction finding: KrakenSDR, with open acquisition and processing software, on Raspberry Pi</li>
<li>BLE: Nordic nRF52840 sniffer with the official firmware and Wireshark</li>
<li>Wi-Fi: Kismet and a monitor-mode adapter</li>
</ul>
<p>None of these bricks is a secret, and that is the point, because the bespoke work stays where it creates value, which is correlating these signals, the local decision that follows and the signed log that attests to it.</p>
<p><em>Next in the series, <a href="/blog/jamming-relays-and-the-flipper-zero">Jamming, Relays and the Flipper Zero</a>, because everything above assumes the spectrum is telling the truth, and we look at what happens when someone is deliberately jamming it or relaying a signal.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Who Is in the Room?</title>
      <link>https://trust-buster.com/blog/who-is-in-the-room</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/who-is-in-the-room</guid>
      <pubDate>Thu, 10 Sep 2026 15:57:40 GMT</pubDate>
      
      <description>Counting radio addresses is not counting people, and one mechanism decides whether you know who is in the room, for enrolled badges, or track nobody at all.</description>
      <content:encoded><![CDATA[<p>A dashboard announcing forty-seven devices detected this morning in a workshop six people walked through is not lying deliberately, because it is counting addresses, and modern Bluetooth and Wi-Fi devices change address every few minutes, which makes naive counting useless and makes one specific mechanism the difference between knowing who is in the room, for the badges you enrolled, and tracking nobody at all.</p>
<p><em>First of four articles on radio detection in industrial environments, and this one covers what you can learn passively, before any spectrum analysis or any anti-jamming logic.</em></p>
<h2 id="three-questions-people-often-conflate">Three questions people often conflate</h2>
<p>Before protocols, a distinction that took us a while to phrase properly, because radio detection covers three needs with different means and different limits. The first is whether this badge is here now, which means listening for a known beacon and tracking its presence, and the difficulty there is not hearing it but telling real proximity from a relayed signal. The second is what is transmitting in this room, which means inventorying everything, known or not. The third is whether someone is blinding us, which means measuring the state of the spectrum itself. This article answers the first, <a href="/blog/what-an-sdr-finds">the next one</a> and <a href="/blog/jamming-relays-and-the-flipper-zero">the one after</a> answer the others, and <a href="/blog/is-radio-monitoring-in-a-workplace-legal">the fourth</a> answers a question none of them can dodge, which is whether you are allowed to do any of it.</p>
<h2 id="what-a-bluetooth-device-says-without-being-asked">What a Bluetooth device says without being asked</h2>
<p>Bluetooth Low Energy works by advertising, because a device that wants to be discovered emits, every few hundred milliseconds, a small packet containing its address, sometimes a name and sometimes service data, and these packets travel in the clear on three dedicated channels that any receiver can listen to without connecting to anything. That is what makes detection easy, since a simple Bluetooth card in listening mode sees smartwatches, earbuds, item trackers, temperature sensors, phones in discovery and the badges we have enrolled, and each packet comes with a measurement, the received signal strength or RSSI, which decreases with distance, and that is what creates the illusion of a distance measurement, and we will come back to why it is a poor one.</p>
<h2 id="the-problem-everyone-eventually-discovers-mac-randomization">The problem everyone eventually discovers: MAC randomization</h2>
<p>A naive system counts addresses and believes it is counting devices, and it is badly wrong, because modern devices, under what the phone makers call MAC randomization, draw a new address roughly every fifteen minutes, so to an outside observer a phone becomes a different device four times an hour, and anyone building a presence system on address counts will overestimate the occupancy of a room by a factor of four to ten and never notice, because the number looks plausible. But the mechanism has a property you can rely on, and it makes all the difference between an enrolled device and a stranger, because each rotating address is derived from a key exchanged once, at pairing, so whoever holds that key can verify that an apparently random address belongs to the device they know, while whoever does not holds only a sequence of unrelated numbers. The consequence is worth stating carefully because it cuts both ways in our favour, since enrolled badges stay tracked despite their address changes, because the key was handed to us deliberately by someone who agreed to be identified, while devices we do not know remain untrackable by construction, since the protocol makes it impossible without the key, whatever our intentions. That is a property of the specification rather than a privacy policy we wrote, and we consider it an advantage rather than an obstacle, and it also means that when someone asks whether our system tracks visitors&#39; phones, the answer we can give is that it cannot, which we find a better answer than a promise that it does not.</p>
<h2 id="entering-and-leaving-the-room-the-single-threshold-trap">Entering and leaving the room: the single-threshold trap</h2>
<p>Deciding that a badge entered when its signal passes a value and left when it drops below produces an unusable system, because the measurement swings ten to twenty decibels when someone turns round, walks behind a machine or sets their badge on a metal bench, so a single threshold generates dozens of entries and exits per hour for a person sitting still in the room, and every one of them is an event someone has to look at. Three mechanisms fix this, and they apply to any system of this kind. The first is hysteresis, which means two thresholds instead of one, so you enter at one level and only leave at a markedly lower one, and the grey zone between produces no event at all. The second is a persistence window, because a badge is declared absent only after several consecutive missed advertising cycles and not on the first lost packet, so a burst of interference does not cut an access in progress, which matters when the consequence of cutting is a machine stopping mid-cycle. The third is temporal smoothing, which means working on a rolling average of recent measurements and never on the instantaneous value. Tuning these three is specific to each room, because a machine shop full of cast iron does not have the same signature as an electronics lab, and neither resembles an outdoor substation, and that is exactly what Observation mode exists to measure, since before anything is armed it produces the real event rate per day and per machine, on your site, with your people, and if the number is not liveable you do not arm that machine, which is a decision and not a failure.</p>
<h2 id="what-wi-fi-adds-and-where-it-stops">What Wi-Fi adds, and where it stops</h2>
<p>A Wi-Fi card in monitor mode, which means passive listening that joins no network, captures a class of frames that encryption does not protect, the management frames, such as access point beacons, client probe requests, association requests and deauthentication frames, and even on a WPA3 network these travel in the clear since they are what negotiates the connection in the first place. That is enough to spot three useful things. The first is an unauthorised access point in the room, such as a cellular box plugged into a cabinet or a phone tethering on a rack, and Bastille has reported a data centre case where a smartphone configured as a hotspot, paired with a dongle attached to a server rack, opened a data path running over the phone&#39;s cellular link that bypassed firewall, DLP and SIEM entirely, with no sophisticated tooling and nothing on the network seeing it. The second is a deauthentication attack, the most common Wi-Fi attack and an easy one to detect, because it shows as an abnormal volume of forced-disconnect frames from a single address, and it is worth noting that it fails against networks that enable protected management frames under 802.11w, which WPA3 does, assuming the estate supports it, which in an industrial setting it frequently does not. The third is an evil twin, an access point announcing a legitimate network name from a room where it has no business being. Where it stops is also worth saying, because for several generations now iOS and Android devices use a different address per network and randomise their probe requests too, so counting Wi-Fi addresses to count people is even less reliable than in Bluetooth, and unlike BLE there is no equivalent of the pairing key to rescue you, which means Wi-Fi listening is good for spotting infrastructure that should not be there and poor for knowing who is present.</p>
<h2 id="what-presence-detection-proves">What presence detection proves</h2>
<table>
<thead>
<tr>
<th>Mechanism</th>
<th>Proven</th>
<th>Not proven</th>
</tr>
</thead>
<tbody><tr>
<td>Receiving a BLE advertisement</td>
<td>An emitter with this identifier transmitted in range</td>
<td>That its legitimate holder is there</td>
</tr>
<tr>
<td>Address resolution via the pairing key</td>
<td>This emitter is the enrolled device</td>
<td>That it was not lent or stolen</td>
</tr>
<tr>
<td>Signal strength with hysteresis</td>
<td>A plausible, sustained radio proximity</td>
<td>A real distance, because the signal can be relayed</td>
</tr>
<tr>
<td>Wi-Fi monitor mode</td>
<td>Abnormal activity or an undeclared access point</td>
<td>A complete inventory of devices present</td>
</tr>
</tbody></table>
<p>The second row is the one that matters operationally, because a badge proves a badge, and someone handing theirs to a colleague because it is quicker defeats every property in this article, so what still stands in the way in that case is the schedule, the persistence requirement, the integrity of the enforcement device, and a log that records the access under the legitimate holder&#39;s name at an hour when that person can say where they were, rather than the badge itself, and that is as close as we get to knowing who is in the room.</p>
<p><em>Next in the series, <a href="/blog/what-an-sdr-finds">What an SDR Finds</a>, about what a software-defined radio effector finds when you look beyond Bluetooth and Wi-Fi, and the structural limit that means it will never find everything.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>The Novel Was Right</title>
      <link>https://trust-buster.com/blog/the-novel-was-right</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/the-novel-was-right</guid>
      <pubDate>Thu, 10 Sep 2026 15:35:01 GMT</pubDate>
      
      <description>Blackout by Marc Elsberg imagined a grid attack in 2012, and we reread the novel that was right about its entry points against the real incidents since.</description>
      <content:encoded><![CDATA[<p>A piece of technical speculative fiction usually ages badly, because its details end up looking quaint, but <em>Blackout</em> by Marc Elsberg has aged the other way, since every real incident since 2012 has made its assumptions a little less bold, and that is why we say here that the novel was right, while looking honestly at where it was wrong and at what that tells us about the equipment nobody keeps an eye on.</p>
<p><em>A lighter article than usual, since we are not discussing standards or architecture but a piece of technical fiction overtaken by the facts, and if you are after a holiday read that will leave you mildly uneasy about your electricity meter, you are in the right place.</em></p>
<h2 id="the-book">The book</h2>
<p><em>Blackout: Tomorrow Will Be Too Late</em> was published in 2012 by the Austrian author Marc Elsberg, has been translated into some fifteen languages and has sold over a million copies, and it is the book that made his name before <em>Zero</em> two years later.</p>
<p>The premise fits in a sentence, because on a winter evening the European power grid begins to fail, and the origin traces back to compromised smart meters, the electricity meters that talk back to the utility, in Italy and Sweden, then Piero Manzano, a former Italian hacker, works out before the authorities do that this is not a fault, and what follows is a race against time as Europe goes dark country by country.</p>
<p>What makes the book interesting to anyone working in our field is less the plot than the choice of attack vectors, because Elsberg does not blow up a power station but goes through the meters, then through the supervision software, the SCADA in the trade&#39;s vocabulary, which in the novel equips most installations, injecting code that triggers false alerts, and he adds a few sabotaged transmission lines so that the grid begins to fall over on its own, and there is no spectacular destruction in this book but misplaced trust in field data, which is the opposite of the Hollywood scenario.</p>
<h2 id="what-was-already-true-when-he-was-writing-from-stuxnet-to-sh">What was already true when he was writing, from Stuxnet to Shodan</h2>
<p>Elsberg invented very little, and he does not pretend otherwise, because two things were already public when he wrote.</p>
<p>Stuxnet, discovered in 2010, had shown that a program could alter the behaviour of industrial controllers, the programmable logic controllers or PLCs, those hardened computers that drive machines, while feeding operators reassuring measurements, so on this point the novel was right before it was even published, because lying to the supervision data, which is its core, was already a documented technique rather than an author&#39;s idea.</p>
<p>Search engines for connected devices had existed since 2009, and Shodan, then Censys and a few others, do not look for pages but for machines, because they continuously probe internet-reachable addresses, note what answers on which port and file the responses in a searchable index. The majority use is legitimate, since researchers measure an estate&#39;s exposure, national agencies warn operators and security teams check what is sticking out of their own perimeter, but these indexes also catalogue industrial equipment that should never have been reachable, and as early as 2011 an <a href="https://www.cl.cam.ac.uk/~fms27/papers/2011-Leverett-industrial.pdf">academic study at Cambridge</a> showed that a map of exposed industrial control systems could be drawn simply by querying these databases, so the finding has been public for fifteen years. A later study made the point sharper, because researchers deliberately exposed four Allen-Bradley controllers on the internet and measured how long it took for them to be indexed, and the answer was nineteen days.</p>
<h2 id="what-reality-has-added-since">What reality has added since</h2>
<p>This is where the fiction gets overtaken.</p>
<p>Ukraine, in December 2015, saw the first power cut publicly confirmed as caused by a cyberattack, with hundreds of thousands of homes without electricity, substations opened remotely and operators watching their cursors move on screen without being able to stop them, and it was repeated the following year with a more polished tool. The <a href="https://nsarchive.gwu.edu/document/15331-sans-and-electricity-information-sharing-and">report</a> notes that the initial phishing dated from March 2015, nine months before the outages, which contrasts with everything else on this list because it is the case that demonstrates patience.</p>
<p>Triton, in 2017, was an attack on a petrochemical site that targeted the safety instrumented system, the controller whose only job is to shut the plant down when something goes wrong, rather than production itself, and crossing that line meant going after the last safety net, which no fiction had thought of. The plant went to a safe state by itself after several controllers detected an anomaly caused by bugs in the malware, so the attack failed on a coding defect of the attacker&#39;s and not on a defence, as the <a href="https://www.ic3.gov/CSA/2022/220325.pdf">FBI notification</a> documents. A physical protection existed all the same, because the Triconex safety controller only accepts a program upload when its key switch, turned by hand on the equipment, is in the PROGRAM position, but it had been left in that position, so the control was not bypassed but disarmed for convenience, which is in our experience by far the most frequent failure mode of the safety devices we come across.</p>
<p>In Bremanger, in Norway, on 7 April 2025, attackers took control of a valve panel at the Risevatnet dam and opened it fully, so that five hundred litres per second flowed for four hours before detection, with no injuries and no structural damage because the flow barely exceeded the structure&#39;s regulatory minimum. Two things make this the emblematic case of the decade in our eyes, because entry was through a control panel reachable on the web and protected by a weak password, and because the structure is not what the word dam suggests, since it generates no electricity, its controls relate to a fish farm and its owner is a small local property company, and in August 2025 the head of the Norwegian police security service publicly attributed the attack to pro-Russian actors, describing it as intended to produce fear and disorder, which is a rare official attribution.</p>
<p>In Aliquippa, in Pennsylvania, in November 2023, a water authority lost control of a booster station, and the controller was a common model, exposed on the internet on its default programming port and protected by a default password or none. According to the joint advisory published by CISA and its partners, at least seventy-five devices were compromised, thirty-four of them in the US water sector, across roughly four waves running into January 2024, with victims in the United States, Israel, the United Kingdom and Ireland, and the attackers went further than the defacement most coverage reported, since they developed custom ladder logic, the programming language of these controllers, for each device type.</p>
<p>And across Europe more broadly, in January 2026 the French national cybersecurity agency <a href="https://www.cert.ssi.gouv.fr/dur/CERTFR-2025-DUR-003/">published a recommendation</a> to operators in the energy and water sectors, which states that several French entities in renewable generation and water management were compromised, that hacktivist groups reached management interfaces of industrial installations exposed on the internet, and that the most successful action stopped a wind farm for several hours with a financial loss to the operator. Three details of that advisory reward slow reading, because the scope covers hydroelectric, wind, photovoltaic and water pumping and treatment installations, because the agency states the activity is generally of a low technical level, with equipment exposed without authentication or behind default passwords and industrial protocols reachable directly from the internet without encryption, and because in some cases these installations are operated by very small companies or by individuals.</p>
<p>The novel imagined brilliant, patient and magnificently equipped attackers, while reality more than a decade later is still often factory passwords on a port documented in the manufacturer&#39;s manual.</p>
<h2 id="what-the-novel-exaggerates-and-why-saying-so-matters">What the novel exaggerates, and why saying so matters</h2>
<p>We would not be consistent with the rest of this blog if we let it be understood that the novel was right about everything, so here are three reservations.</p>
<p>The domino effect is too fast and too total, because European grids have load-shedding mechanisms, staged restoration and manual operation that do not yield to a single line of code, so a continental collapse in a few hours is narrative licence.</p>
<p>The software monoculture is a plot device, since the novel assumes one supervision package equips most installations, while reality is a patchwork of vendors, versions and local improvisation, which is a nightmare to secure but also a good protection against a single universal attack.</p>
<p>And there is a correction we owe our own side, because the Oldsmar water treatment incident in Florida, in 2021, circulated for a long time as the perfect example of remote takeover, while later analysis concluded it was more likely an operator error on a shared workstation than an intrusion, and it still appears in vendor presentations. We would rather remove it from the list than pad the file, but it is worth dwelling on because it illustrates two things this article is really about, since a version circulates and the correction never entirely catches up with it, and since a well-meaning workaround of a control felt to be cumbersome also leads to handling errors with serious consequences.</p>
<h2 id="the-novel-was-right-but-not-where-one-expects">The novel was right, but not where one expects</h2>
<p>Closing the book, the temptation is to conclude that large infrastructure needs better protection, and that is true, and largely under way.</p>
<p>The more interesting lesson lies elsewhere, and it emerges from comparing the novel with the real incidents, because in almost every documented case the entry point was not the central system but a valve panel on a fish farm, a cellular modem on a small flood-control structure, a controller on a booster station serving three thousand people, a wind farm run by a very small company, or equipment nobody considered critical, installed by a contractor, never inventoried, never updated and reachable from anywhere. That is precisely the estate our other articles are about, too physical for IT, too digital for maintenance and on nobody&#39;s list, and this is where the novel was right without quite knowing it, because it looked for its weaknesses in the right place, at the edges, even if it lent them attackers who were far too brilliant.</p>
<p>One honest note to close this point, because nothing we build solves the problem of a controller exposed on the internet, since that is network segmentation and remote access management and there are very good people doing it. Our business is the other path, <a href="/blog/remote-access-and-physical-access-two-different-problems">someone standing physically in front of the machine</a>, and both paths exist without overlapping, so it seems prudent to us to be wary of anyone claiming to close both with a single product, ourselves included.</p>
<h2 id="so-is-it-worth-reading">So, is it worth reading?</h2>
<p>Yes, first because it is a good thriller, which is no small thing for a book that explains the workings of a transmission grid in passing, then because it is the best popularisation tool we have found so far for making cascading dependency understood by a non-technical audience, and getting a director to read three chapters is often more effective than a forty-slide presentation, and finally because it ages in the opposite direction to most speculative fiction, since each passing year makes one of its assumptions a little less bold. Saying the novel was right is not a very reassuring compliment, but it is the one the book deserves.</p>
]]></content:encoded>
    </item>
    <item>
      <title>Purdue Diligence: The Purdue Model in Ten Minutes</title>
      <link>https://trust-buster.com/blog/purdue-diligence</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/purdue-diligence</guid>
      <pubDate>Thu, 10 Sep 2026 14:15:49 GMT</pubDate>
      
      <description>The Purdue model sorts industrial systems into six levels, and here is where it comes from, what each level holds and how the standards lean on it.</description>
      <content:encoded><![CDATA[<p>The Purdue model organises industrial systems into six levels numbered 0 to 5, to which practice has added a demilitarised zone inserted at level 3.5, and since it was published in 1992 and remains the common language of the shop floor and the architecture board alike, knowing which level each of your systems sits at, and who is allowed to talk to whom, is the starting point of any discussion about access to your machines.</p>
<p><em>We cite it because it is frequently referred to in the literature, and sometimes used as a sales argument, without necessarily matching its definition, so here is the short version for anyone arriving in the industrial world from business IT and cybersecurity, ahead of the two following articles in this series, which look at <a href="/blog/with-all-purdue-respect">what the model does not see</a> and at <a href="/blog/long-purdue">the link it is missing</a>.</em></p>
<h2 id="where-it-comes-from-and-what-it-was-aiming-at">Where it comes from, and what it was aiming at</h2>
<p>In 1992, Theodore J. Williams and the Purdue University consortium published a reference model for computer-integrated manufacturing, the Purdue Enterprise Reference Architecture or PERA, and the context is the computerisation of factories, long before Industry 4.0, because management systems were beginning to be connected to production systems and nobody quite knew how to describe the whole.</p>
<p>The idea is to organise the environment into distinct functional levels, each with its own role, its own systems and its own stakeholders, and it is not a security architecture, since the word barely appears in the original project, but a data flow model intended to bring order to a factory and to get people talking to one another who did not share the same vocabulary.</p>
<p>That origin explains both its longevity and its limits, and above all it explains why the Purdue model was turned into a security framework through usage rather than by design, because it was there, its hierarchy was convenient and it offered an obvious place to put firewalls.</p>
<h2 id="the-six-levels-of-the-purdue-model">The six levels of the Purdue model</h2>
<h3 id="level-0-the-physical-process">Level 0, the physical process</h3>
<p>This is what produces, heats, turns and measures, so motors, valves, pumps, actuators and temperature, pressure or position sensors, and it is the world of contact with matter, the only level where energy is exerted on it, and therefore the one where any error made elsewhere ends up translating into physical risk. It is the level we seek to protect without residing in it, as we explain in <a href="/blog/long-purdue">the third article of this series</a>.</p>
<h3 id="level-1-basic-control">Level 1, basic control</h3>
<p>This is the equipment that drives level 0, that is to say programmable logic controllers, the hardened computers that run the control logic, field computers, regulators and intelligent electronic devices, and this is where the thing that decides a valve opens when a temperature exceeds a threshold resides.</p>
<p>It is also, often, the oldest level of the installed base, because a PLC installed in 2003 does exactly what it is asked to do and has no reason to be replaced, but it does not necessarily incorporate access control, since the security asked of it at the time was that of the service it provides, which is to do its job without wear and without fault, and not that of knowing who reprograms it.</p>
<h3 id="level-2-area-supervision">Level 2, area supervision</h3>
<p>These are the human-machine interfaces, the operator workstations and the control systems of a workshop or a line, where the operator sees the state of the process and acts on it, and technically we are already in the world of the general-purpose operating system, often a version of Windows whose update would mean stopping production.</p>
<h3 id="level-3-site-operations">Level 3, site operations</h3>
<p>This is production management at plant scale, with manufacturing execution systems, data historians, batch management and scheduling, and it is the last level considered to belong to the industrial world, with industrial IT, machine-to-machine networking and the first mechanisms for identifying users.</p>
<h3 id="level-3-5-the-industrial-demilitarised-zone-or-idmz">Level 3.5, the industrial demilitarised zone or IDMZ</h3>
<p>This level did not exist in the original model, because it was added by practice when the plant and the enterprise had to be made to talk to each other without being put in direct contact, and it is a buffer zone, the industrial DMZ, where industrial systems deposit data that management systems come to collect, with no flow crossing from end to end. It houses relay servers, replicated historians and sometimes unidirectional gateways that make the return path physically impossible, and this half-level bears a large share of the model&#39;s security burden, which explains why it is talked about so much.</p>
<h3 id="level-4-site-planning">Level 4, site planning</h3>
<p>These are enterprise resource planning, logistics, resource planning and the site&#39;s business applications, and by this point we have crossed over to business IT, with network equipment that we know how to monitor.</p>
<h3 id="level-5-the-enterprise">Level 5, the enterprise</h3>
<p>This is the group network, shared services, the internet connection and enterprise applications, in other words the world a conventional chief information security officer knows.</p>
<h2 id="the-operating-principle">The operating principle</h2>
<p>The model&#39;s implicit rule, as it is usually read, fits in a single sentence, because a level only talks to its immediate neighbours, so level 4 does not address a level 1 PLC directly but goes through level 3, which goes through level 2, and each boundary becomes a place where a control can be put, where traffic can be filtered, logged or refused.</p>
<p>It is this property that made the fortune of the Purdue model in security, because it turns a factory into a succession of compartments and gives a simple answer to a difficult question, which is where the firewall should go, and it is also why the first use of the model is an inventory, since you have to know what sits at each level before knowing where to filter.</p>
<h2 id="how-the-frameworks-use-it">How the frameworks use it</h2>
<p>The model did not remain on its own, since it has been absorbed by more or less everything that matters in industrial security, after the ISA-95 standard, which describes the integration between enterprise and control systems, had already taken up its functional hierarchy.</p>
<p>IEC 62443 defines its zones, conduits and trust boundaries on the basis of Purdue, and a zone groups together assets sharing the same protection requirements, while a conduit is a controlled communication path between two zones, and part 3-2 of the standard lays down a strong rule, because all communication between zones passes through an explicitly defined and monitored conduit, which we detail in our reading of the standard.</p>
<p>NIST SP 800-82, the American guide to operational technology security, relies on it to describe its layered defence in depth, and the reference architectures of the major vendors derive directly from it, such as Cisco and Rockwell&#39;s converged architecture or Schneider Electric&#39;s secure reference architecture, while the SANS SCADA reference model overlays on the Purdue hierarchy the places to put firewalls, proxies, detection sensors and jump hosts.</p>
<p>And the &quot;Zero Trust for Operational Technology&quot; doctrine published by the US Department of Defense in November 2025 explicitly recognises the Purdue model and IEC 62443 as frameworks that remain authoritative for classifying OT systems, even though it then chooses another grid, which we discuss in the next article.</p>
<h2 id="what-to-take-away">What to take away</h2>
<p>Three things seem to us worth keeping if you are only to keep three.</p>
<p>It is a descriptive model before it is a prescriptive one, because its primary value is to say what exists, which system sits where and how data should flow, and the rest is a layer of usage.</p>
<p>It is the common language, because when a security architect says &quot;the 3.5&quot;, an automation engineer immediately understands what they are talking about, and in a field where IT and automation understand each other poorly that is an underestimated quality.</p>
<p>And it describes systems, not people, since each of its levels relates one system to another system, which is consistent with what was asked of it in 1992 and is the starting point for the rest of this series.</p>
<p><em>Next in the series: <a href="/blog/with-all-purdue-respect">the criticisms levelled at the Purdue model in recent years</a>, because the industrial internet of things, the cloud and remote access short-circuit its hierarchy, but they all share a common point that is rarely noticed, and that is the subject of the second article.</em></p>
<p><em>Sources:</em></p>
<ul>
<li>Williams, T. J. &quot;The Purdue enterprise reference architecture&quot;. <em>Computers in Industry</em>, vol. 24, no. 2-3, 1994, pp. 141-158. DOI: 10.1016/0166-3615(94)90017-5.</li>
<li>Williams, T. J. <em>The Purdue Enterprise Reference Architecture</em>. Instrument Society of America, 1992.</li>
<li>International Electrotechnical Commission. <a href="https://webstore.iec.ch"><em>IEC 62443-3-2:2020, Security risk assessment for system design</em></a>. Geneva: IEC, 2020.</li>
<li>International Electrotechnical Commission. <a href="https://webstore.iec.ch"><em>IEC 62443-3-3:2013, System security requirements and security levels</em></a>. Geneva: IEC, 2013.</li>
<li>National Institute of Standards and Technology. <a href="https://csrc.nist.gov/pubs/sp/800/82/r3/final"><em>SP 800-82 Rev. 3, Guide to Operational Technology (OT) Security</em></a>. September 2023.</li>
<li>Department of War, Chief Information Officer. <a href="https://dodcio.defense.gov/Portals/0/Documents/Library/ZT-OperationalTechnologyActivitiesOutcomes.pdf"><em>Zero Trust for Operational Technology Activities and Outcomes</em></a>. 18 November 2025.</li>
<li>Cisco Systems and Rockwell Automation. <a href="https://www.cisco.com/c/en/us/solutions/design-zone.html"><em>Converged Plantwide Ethernet (CPwE) Design and Implementation Guide</em></a>.</li>
</ul>
]]></content:encoded>
    </item>
    <item>
      <title>Security Principles in OT</title>
      <link>https://trust-buster.com/blog/security-principles-in-ot</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/security-principles-in-ot</guid>
      <pubDate>Mon, 24 Aug 2026 14:42:17 GMT</pubDate>
      
      <description>The security principles in OT behind a workshop access control system, tested against twelve attacks, with what each mechanism proves and what it does not.</description>
      <content:encoded><![CDATA[<p><em>People usually ask us whether &quot;it&#39;s encrypted&quot;, and the question is a fair one, but it is not enough, because the security principles in OT do not boil down to encryption and there are plenty of ways to defeat an access control system fitted to electrical equipment.</em></p>
<p><em>So this article starts with a precise statement of what we do to secure the whole process, describes who the solution is meant to hold against, and then works through the concrete ways someone might start a machine they have no right to start.</em></p>
<h2 id="a-security-walkthrough-of-a-workshop-access-control-system">A security walkthrough of a workshop access control system</h2>
<h3 id="the-claim-we-make">The claim we make</h3>
<p>Here is what we sell with the first solution Ops-Devices has developed, in the context of machines to be switched on or off according to who is authorised:</p>
<p><strong>No protected equipment can be energised unless an enrolled badge is present, and only for as long as it stays present, with the system checking that its holder is acting within the time slot granted to them, on the machines they are authorised for, and that the switching device is intact and reachable. Every attempt, successful or not, must leave a trace timestamped to workshop time.</strong></p>
<h3 id="the-setup">The setup</h3>
<p>Four machines in one room, with no modification whatsoever to the protected equipment, a radio <strong>badge</strong> carried by the person intervening, a standalone <strong>decision unit</strong> within range of the zone, a <strong>head-end module</strong> in the cabinet that feeds and meters each channel, four <strong>controlled outlets</strong> on the machines&#39; power supply, and the supervisor&#39;s <strong>mobile app</strong>, which shuttles between the machine room and the offices (keeping a daily link with the administration server, for instance).</p>
<p>The unit has no network access at all and every decision is taken locally, which has a direct consequence for the demonstration, because cutting the internet changes absolutely nothing in how the system behaves, but the flip side is that alerts stay confined to the plant room. The access rules therefore have to be right, and sometimes strict, since no remote confirmation from a supervisor is available.</p>
<h3 id="the-threat-model-or-the-adversary-we-designed-against">The threat model, or the adversary we designed against</h3>
<p>We are placing ourselves against someone who <strong>has legitimate or tolerated physical access to the zone</strong>, a subcontractor, a temp, a service provider, a visitor left alone, who has ordinary tools and a few minutes, and whose motivation is <strong>convenience</strong>, that is to get a machine running and move on, working around a procedure they find cumbersome, to save time.</p>
<p>We are not placing ourselves against an adversary who would set up a radio relay, analyse an advertisement and clone it, or swap a module for a hardware clone. Those threats are real, they belong to a different budget and a different system, one we have designed, and they are the subject of other articles, including the one on <a href="/blog/jamming-relays-and-the-flipper-zero">jamming, relays and the Flipper Zero</a>.</p>
<h2 id="the-foundation-of-our-security-principles-in-ot-three-bounda">The foundation of our security principles in OT: three boundaries and one fallback position</h2>
<p>The claim above rests on three distinct boundaries, none of which can do the work of the other two, and on a default behaviour that makes every failure harmless.</p>
<table>
<thead>
<tr>
<th>Boundary</th>
<th>The question it answers</th>
</tr>
</thead>
<tbody><tr>
<td>Human → decision</td>
<td><em>Who is here, and do they have the right, now?</em></td>
</tr>
<tr>
<td>Decision → enforcement</td>
<td><em>Does this order come from the authority, and is it valid?</em></td>
</tr>
<tr>
<td>Enforcement → machine</td>
<td><em>What can this device do, and what must it never do?</em></td>
</tr>
</tbody></table>
<p>There is a familiar parallel here with the keyed inserts on UPS outlets, because every sysadmin has chased colleagues who plug a coffee machine or a bike battery into the red sockets, since &quot;we&#39;ve got our own priorities anyway&quot;. So we replace the plastic keying with equipment that cannot be prised open with a biro barrel, and more to the point we avoid the production halted, the machining or 3D printing hours lost, and the servers shut down mid-experiment.</p>
<p><strong>The fallback position is denial.</strong> Outlets stay off until a positive decision has been taken, not the other way round, which is fail-secure behaviour, so a unit failure, a lost link, an unreadable badge, a missing schedule or a jammed radio all return to the same de-energised state.</p>
<p>That choice carries a real operating cost, since the system has to work for production to run, and we accept it, because an access control that fails open is not a degraded control, it is an absence of control with a green light on. Security here is physical (unplug a cable, swap two devices, spoof the sensors) before it is ever a matter of protocols.</p>
<h2 id="the-demonstration-attack-by-attack">The demonstration, attack by attack</h2>
<p>Each attack below is described with what stands in the way, why it holds and what remains open, because that is how we prefer to present security principles in OT, by what they concretely prevent rather than by the list of algorithms involved.</p>
<h3 id="1-walk-up-to-the-machine-and-press-the-button">1. Walk up to the machine and press the button</h3>
<p><strong>The attack.</strong> The most common by far, because nobody hacks anything, you stand in front of a connected outlet and press the switch. We had assumed everything would be driven remotely, but even on budget hardware there is always a physical alternative.</p>
<p><strong>What stands in the way.</strong> On the outlets we recommend, the physical button is <strong>decoupled from the relay</strong> by configuration, so pressing it no longer commands the output and only the decision layer can energise. Simple, but necessary, and it justifies spending €30 instead of €10.</p>
<p><strong>Why it holds.</strong> Because the link between button and relay no longer exists in the outlet&#39;s firmware, which we have kept control of, rather than merely being refused after the fact by a software rule. There is nothing to bypass since the function is gone, and without this setting everything else would be decorative, because anyone standing in front of the outlet would take back control of the equipment.</p>
<p><strong>What remains.</strong> A factory reset of the module would restore its original behaviour, but that takes prolonged physical access and announces itself, because the module reappears under its default identity and stops answering the unit, which triggers the mechanism described in point 2, a second line of defence in other words.</p>
<h3 id="2-unplug-the-outlet-and-plug-the-machine-in-elsewhere">2. Unplug the outlet and plug the machine in elsewhere</h3>
<p><strong>The attack.</strong> The common-sense workaround, and the one we most often see overlooked in competing systems, because if the outlet refuses, you remove it, and the answer is not more glue.</p>
<p><strong>What stands in the way.</strong> Each outlet is fed by a channel of the head-end module in the electrical cabinet, and if the unit loses the link to an outlet for several consecutive cycles, it <strong>cuts the channel feeding it</strong> and records the event in the log as critical, but only after trying to find it again, so that ordinary radio interference does not trigger a reaction. We selected components that are reliable but affordable, the aim being to scale up and equip every outlet.</p>
<p><strong>Why it holds.</strong> Because the cut happens upstream of the device that was removed, at a point the attacker is not holding while they manipulate the outlet, so the simplest workaround costs them a power cut and leaves a trace.</p>
<p><strong>What remains.</strong> The protection only covers what the head-end module feeds, so plugging the machine into a wall socket outside the perimeter falls outside the system entirely, and is a matter for the electrical installation rather than for access control.</p>
<h3 id="3-replay-a-captured-order">3. Replay a captured order</h3>
<p><strong>The attack.</strong> Record the radio exchanges while a legitimate holder obtains access, then replay them later for the same effect, which is the classic replay attack against any system that merely encrypts. Two years ago it was a heavy workaround requiring real knowledge, but since the spread of the Flipper Zero and similar devices, the move is within reach of an amateur with no particular training.</p>
<p><strong>What stands in the way.</strong> When the app connects to the unit, the unit draws a <strong>thirty-two byte nonce</strong> (an unpredictable, single-use number) and expects in return an attestation signed by the authority, together with that nonce signed by the device&#39;s key, and the nonce is <strong>consumed no matter what</strong>, whether the attempt succeeds or fails. This is our own contribution, because it did not exist in the off-the-shelf equipment.</p>
<p>Three further checks cover rights updates, since an update addressed to a different unit is refused, an update whose sequence number is not higher than the last one applied is ignored, and the list of authorised holders is <strong>replaced wholesale</strong>, never merged.</p>
<p><strong>Why it holds.</strong> Because replay does not merely become ineffective, it becomes <strong>impossible to repeat</strong> and it is detectable, since every attempt burns the current nonce and every stale packet meets a sequence number that does not go backwards. Without wholesale replacement, an attacker could replay an old packet to resurrect a revoked holder, and that is precisely the flaw this innocuous-looking detail closes.</p>
<p>The refusal also comes <strong>with a reason</strong>, because the unit answers by naming it, no nonce, unrecognised authority, invalid proof of possession, expired attestation or holder not on the list, so that a rejection is distinguishable from a fault, both in the log and on the supervisor&#39;s screen. We have also worked out a colour code that lets you walk into a room and see at a glance what is cut, what is live, and what has been tampered with.</p>
<h3 id="4-impersonate-the-authority">4. Impersonate the authority</h3>
<p><strong>The attack.</strong> Forge a rights update to add yourself to the list of authorised holders, or to widen a time slot.</p>
<p><strong>What stands in the way.</strong> The unit verifies the authority&#39;s signature using a public key <strong>embedded in its firmware</strong>, which the sender of the message can neither supply nor replace, and verification is performed on the bytes exactly as received, <strong>before</strong> any interpretation.</p>
<p><strong>Why it holds.</strong> The order of operations is not a detail, because verifying the received bytes and then parsing them, rather than the reverse, eliminates a whole class of flaws in which the message as written differs by a byte from what was signed, and tampering with the content breaks the signature as long as the authority&#39;s private key stays secret, which brings us back to point 12.</p>
<p>As for anchoring the key in firmware, it means a software update is required to change it, and that is a drawback we much prefer to the risk of an attacker replacing the authority itself. Over-the-air update mechanisms, with no need to open the Cube, our on-site decision unit (<a href="/trust-buster-solution/cube-and-satellites">overview</a>), or its satellites, are now widely available on a great many components, and we always select parts to which standard IT security practice applies, so that we can concentrate on the security principles in OT that have no equivalent on that side.</p>
<p><strong>What remains.</strong> This anchoring is only worth something if the component cannot be reprogrammed, so hardware secure boot and encryption of program memory are the logical next step.</p>
<h3 id="5-make-it-look-as-though-verification-is-working">5. Make it look as though verification is working</h3>
<p><strong>The attack.</strong> This one is not carried out by an adversary, and that is what makes it formidable, because <strong>a broken cryptographic check is silent</strong>. Either it refuses everything, which is noticed immediately, or, far worse, it accepts everything, and no functional test of the product can tell a valid signature from a broken checking routine that always answers yes.</p>
<p><strong>What stands in the way.</strong> At every start-up, the unit runs a self-test against published reference vectors, checking that it <strong>accepts</strong> an authentic signature and <strong>rejects</strong> a forged signature and an altered message alike, and data decoding and expiry-date parsing are tested the same way, edge cases included.</p>
<p><strong>Why it holds.</strong> Because the test fails safe, since if the self-test does not pass, the unit <strong>refuses to open its administration channel</strong> rather than serving it without checks, so a device whose cryptography is broken does not merely report the fact, it stops accepting orders.</p>
<h3 id="6-wind-the-clock-back-to-open-a-time-slot">6. Wind the clock back to open a time slot</h3>
<p><strong>The attack.</strong> The schedule authorises a holder from seven in the morning to ten at night, it is eleven, and all you need is to convince the unit that it is three in the afternoon.</p>
<p><strong>What stands in the way.</strong> Setting the time is treated as a <strong>mutating operation</strong>, on the same footing as a rights update, so it requires an authenticated session, and without one the request is refused with its reason.</p>
<p><strong>Why it holds.</strong> Because the clock is not a technical convenience but a security parameter, since it governs time slots, attestation expiry and log timestamping, and leaving it freely modifiable would have offered a way around everything else, and made it possible to date anything at will. In the OT world this is a perennial problem, because there is no counting the systems running twelve hours out, still on winter time in mid-summer, or unaware of public holidays, and it is often one of the first reasons a system gets abandoned, particularly in BMS and HVAC.</p>
<p><strong>What remains.</strong> One edge case is worth knowing about, because at the very first start-up, before a trusted time has been supplied, attestation expiry cannot be evaluated, and the unit lets that specific check through, the authority&#39;s signature still being required, and <strong>the exception is written to the log</strong>. It is a deliberate trade-off, visible at audit, and tightening it is on the register, subject to your future requirements.</p>
<h3 id="7-delete-an-inconvenient-line-from-the-log">7. Delete an inconvenient line from the log</h3>
<p><strong>The attack.</strong> Tamper with the system, obtain access (or not, that is what we are claiming), then make the trace disappear, or, more subtly, use the shuttle to lose a report along the way, because a mobile app collects the local logs and carries them to an area with network coverage, so cutting off the app cuts off the administration server&#39;s view.</p>
<p><strong>What stands in the way.</strong> Purging the log requires an authenticated session, and the unit refuses to acknowledge entries that do not yet exist, and each session also carries an identifier drawn at start-up, which lets the app detect that an acknowledgement comes from an earlier session and decline to relay it.</p>
<p>More fundamentally, nothing counts as transmitted until the recipient has confirmed it, because an entry is only purged from the unit once the authority has recorded it, and the transfer log shown to the operator adds a line only after the recipient&#39;s acknowledgement, never after the send.</p>
<p><strong>Why it holds, partially.</strong> These mechanisms prevent accidental loss and cross-session mismatch, which are the real causes of misleading logs, but they do not prove that no entry was removed along the way, and we have devoted an article to what a proof log proves and what it does not.</p>
<h3 id="8-copy-the-badge-s-advertisement">8. Copy the badge&#39;s advertisement</h3>
<p><strong>The attack.</strong> Listen to a badge&#39;s advertisement, reproduce it, and be detected as its holder, or simply hand your badge to a colleague, because &quot;it&#39;s easier, and what matters is getting the job done, right?&quot;</p>
<p><strong>What stands in the way, at badge level.</strong> Nothing, because the advertisement travels in the clear, it is identical every time, and anyone listening can reproduce it. So we <strong>never</strong> present the badge as an authentication factor, since it answers the question &quot;is a holder claiming this identifier nearby?&quot;, and no other.</p>
<p><strong>What stands in the way after that.</strong> The badge is not the barrier, it is the trigger, and still standing in the way are the impersonated holder&#39;s <strong>schedule</strong>, the <strong>persistence</strong> of presence (access is cut as soon as the advertisement disappears), the <strong>integrity of the switching device</strong>, and the <strong>log</strong>, which will record the access under the legitimate holder&#39;s name, at an hour when that person can say where they were.</p>
<p><strong>Why this is acceptable here.</strong> Because an adversary who analyses an advertisement format in order to reproduce it has left the opportunist profile behind, and the measure of risk is not &quot;is it feasible?&quot;, since it is, but &quot;does it match someone who actually turns up in this workshop?&quot;. That is a judgement, it is revisable, and it changes in the contexts covered by another article on the most sensitive cases.</p>
<p><strong>What remains.</strong> The weakest point of the assembly, identified as such, can be addressed without changing hardware, by using an advertisement payload that varies with each transmission, and for those who want it there are excellent badges we could integrate. One can also pair the badge with a secure ID card presented to the Cube over NFC, though that runs into the lending problem again, in which case you may as well use the mobile app confirmation we have already integrated alongside the trigger badge.</p>
<h3 id="9-relay-a-legitimate-holder-s-presence">9. Relay a legitimate holder&#39;s presence</h3>
<p><strong>The attack.</strong> Retransmit the conversation between a holder who has stayed in the locker room and the target machine, manufacturing a perfectly authentic presence, because here the badge is not stolen, a bridge is built between the two, and this is our friend the Flipper Zero again, which opened the Pandora&#39;s box of cyber-physical system security.</p>
<p><strong>What stands in the way.</strong> Nothing cryptographic, and it needs saying plainly, because no signature sees this kind of attack since no message is forged. What is forged is proximity, and proximity is not a property you can sign.</p>
<p>What does raise the cost is that presence has to be <strong>maintained</strong> throughout the period of use, that the holder&#39;s time slot has to be open at the same moment, and that the access is logged under their name.</p>
<p><strong>What remains.</strong> A real, structural vulnerability, and one that only time-of-flight distance measurement closes. We do not claim to cover it in our current solution, but we can walk you through UWB-based approaches (ultra-wideband radio, which measures a distance from the signal&#39;s time of flight) for a v2, if you bring us the requirement and the budget that goes with it.</p>
<h3 id="10-steal-an-operator-s-phone">10. Steal an operator&#39;s phone</h3>
<p><strong>The attack.</strong> Take the device carrying the attestation and use it to push false settings or purge a log.</p>
<p><strong>What stands in the way.</strong> The app requires biometric validation before sensitive operations, and always in emergency mode, the device attestation has a limited lifetime, and revoking a device takes effect as soon as the first rights update is applied, the holder in session being logically disconnected there and then.</p>
<p><strong>Why it holds.</strong> Because the phone decides nothing, since it carries objects signed by the authority and verified by the unit, so stealing it grants the right to <strong>carry</strong>, not the right to <strong>issue</strong>.</p>
<p><strong>What remains.</strong> Revocation only propagates the next time an authorised device passes by, because that is the price of running without a network. And the device key, while it sits in protected storage, is not yet generated inside a hardware enclave, so it is protected from the rest of the device without being non-extractable, and at that point we are back in ordinary IT security territory.</p>
<p><strong>Break-glass, a sealed and traced way out.</strong> A designated supervisor can force access in an emergency through an offline procedure, with a single-use code, printed and sealed in a tamper-evident envelope kept in each zone, scanned and then validated by biometrics on their phone. The code is consumed on use, so an opened envelope is an envelope to replace, and the access is granted immediately, timestamped, signed and reported at the next upload, with an explicit obligation to recommission the system. We never place this code permanently on or inside a piece of equipment, because that would make the bypass permanently available to the very person the system is installed against.</p>
<h3 id="11-cut-power-to-the-unit-or-jam-the-radio">11. Cut power to the unit, or jam the radio</h3>
<p><strong>The attack.</strong> Rather than trying to obtain access, bring the system down and hope it opens.</p>
<p><strong>What stands in the way.</strong> The fallback position, because a Cube with no power commands nothing and the outlets stay in the refusing state, and a jammed radio makes presence disappear, which cuts. There is no path, in what we have designed and tested, by which degradation opens an access that was not intended.</p>
<p><strong>Why it holds.</strong> Because the direction of failure was chosen at design time rather than suffered, and among the security principles in OT we apply, it is the one we verify first.</p>
<p><strong>What remains.</strong> The flip side, which has to be stated to the operator, is that this system is <strong>deniable of service</strong>, because someone who wants to stop production rather than gain access will manage it by jamming the band, and the answer to that is not technical but organisational, with a derogation procedure, an emergency mode and monitoring of radio availability. At that point we are no longer in the security of the system itself but up against a pair of bolt cutters, which is resolved by building access control and sound HR management.</p>
<h3 id="12-attack-the-administration-server">12. Attack the administration server</h3>
<p><strong>The attack.</strong> Do not fight the field at all, but go for the key that signs the rights distributed across an entire estate.</p>
<p><strong>What stands in the way.</strong> Every signed object carries the identifier of the key that signed it, which makes rotation possible without reprovisioning every device, and the server <strong>refuses to sign</strong> if its key is absent, rather than improvising a volatile key that would silently invalidate everything already distributed.</p>
<p><strong>What remains.</strong> This is today the most sensitive point in the chain, and it would be dishonest to present it otherwise, because compromising this key means compromising the entire estate. Keeping it offline, in a dedicated hardware module, with a documented ceremony, is the measure that closes the risk, but here again this is IT security, addressed by proven means.</p>
<h2 id="what-each-mechanism-proves-and-what-it-does-not">What each mechanism proves, and what it does not</h2>
<p>This is the table we consider most useful to keep in view, because a solid system, by the security principles in OT we hold to, is not the one that stacks up the most mechanisms, but the one in which every mechanism is used for what it actually proves.</p>
<table>
<thead>
<tr>
<th>Mechanism</th>
<th>What it proves</th>
<th>What it does not prove</th>
</tr>
</thead>
<tbody><tr>
<td>Badge advertisement</td>
<td>That a transmitter carrying this identifier is nearby</td>
<td>Identity, because the payload is copyable</td>
</tr>
<tr>
<td>Signal strength filter and hold window</td>
<td>A plausible, sustained radio proximity</td>
<td>A real distance, because the signal can be relayed</td>
</tr>
<tr>
<td>Single-use nonce</td>
<td>That the exchange is fresh and not replayed</td>
<td>That the other party is physically present</td>
</tr>
<tr>
<td>Authority signature</td>
<td>The origin and integrity of distributed rights</td>
<td>That the authority key is well guarded</td>
</tr>
<tr>
<td>Increasing sequence and bound recipient</td>
<td>That a stale or diverted order does not apply</td>
<td>No particular reservation</td>
</tr>
<tr>
<td>Start-up self-test</td>
<td>That verification actually works</td>
<td>That the firmware has not been replaced</td>
</tr>
<tr>
<td>Decoupled button</td>
<td>That the local gesture no longer commands the relay</td>
<td>That a factory reset is impossible</td>
</tr>
<tr>
<td>Device monitoring</td>
<td>That a removal is detected and traced</td>
<td>That the machine is not replugged outside the perimeter</td>
</tr>
<tr>
<td>Log carried by shuttle</td>
<td>What the system decided, and when</td>
<td>That no entry was removed along the way</td>
</tr>
<tr>
<td>Fallback position at denial</td>
<td>That no failure opens an access</td>
<td>That the system stays available</td>
</tr>
</tbody></table>
<h2 id="how-to-verify-this-rather-than-take-our-word-for-it">How to verify this rather than take our word for it</h2>
<h3 id="the-declared-encryption-problem">The declared-encryption problem</h3>
<p>In this field there is a gap between what datasheets announce and what equipment does, and that gap is not always bad faith, because a cryptographic suite can be declared in a data format without being used on the channel, a pairing mode can be supported without being required, and a value can be readable without authentication while the product is sold as encrypted.</p>
<p>We chose to treat this as a testing problem rather than a communication one, that is to state the properties the link ought to hold, measure them one by one, and make the ones that fail do so loudly, because security principles in OT that are not measured remain intentions. That test bench is what produced the register above, not the other way round, and our commissioning support therefore aims to reproduce these measurements on your own installation before it goes live.</p>
<h2 id="why-this-is-enough-in-the-vast-majority-of-workshops">Why this is enough in the vast majority of workshops</h2>
<p>Take the twelve attacks again. Seven are closed by construction, the button, the replayed order, the impersonated authority, the broken verification, the wound-back clock, the stolen phone and the induced failure, one is detected and traced pending being blocked, the removed outlet, and four remain open in whole or in part.</p>
<p>And those four (copying an advertisement after analysing it, setting up a relay, losing a log entry along the way, reaching the administration server&#39;s key) have one thing in common, because <strong>they all require leaving the opportunist profile behind</strong>, since they assume hardware, time, intent, and most often a return to the scene.</p>
<p>This is where our judgement distinguishes two solutions, and we present them to you as such. In a workshop, a machine room or a warehouse where the realistic threat is a shortcut taken by people who are legitimate but not authorised, it is enough that every shortcut costs more than the normal procedure, and that is the case our solution is built for. On a site where someone is willing to pay the financial and technical price of getting in, that reasoning no longer holds, and a different system is needed, with hardware, procedural and software measures matched to the potential harm.</p>
<p>The rest is a question of proportion, with four off-the-shelf modules, no intervention on the equipment, no requalification of the process, no network dependency, and a twelve-line register you can put on the table, and that is the concrete form the security principles in OT take, for us, in a workshop. That is what we can already put in place for you today.</p>
]]></content:encoded>
    </item>
    <item>
      <title>OT Retrofit: Addons instead of Upgrades</title>
      <link>https://trust-buster.com/blog/ot-retrofit-addons-instead-of-upgrades</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/ot-retrofit-addons-instead-of-upgrades</guid>
      <pubDate>Mon, 24 Aug 2026 08:54:05 GMT</pubDate>
      
      <description>Why, for the OT retrofit of an industrial machine, we preferred adding units at its boundary to modifying its core, and which building blocks we rely on.</description>
      <content:encoded><![CDATA[<p><em>We have run into the question of &quot;securing&quot; an existing industrial machine often enough, which the trade calls OT retrofit, and the first idea that comes to mind is to look for whatever the manufacturer provides for changing a module or reflashing the firmware on its boards.</em></p>
<p><em>We ended up making a different choice, because once you add up the effects on cost, compliance and security, it seemed the most reasonable one, and here is the reasoning that took us there.</em></p>
<h2 id="why-an-industrial-machine-is-not-managed-like-a-server">Why an industrial machine is not managed like a server</h2>
<h3 id="equipment-built-to-last-fifteen-or-more-years">Equipment built to last fifteen or more years</h3>
<p>Where business IT gets patched, rebooted and replaced every few years, a PLC (programmable logic controller) from the 2000s, an industrial chiller or a hydraulic press was designed to run for over fifteen years, and they were designed at a time when nobody imagined that strong authentication would one day have to be bolted onto them.</p>
<h3 id="no-natural-foothold-for-cybersecurity">No natural foothold for cybersecurity</h3>
<p>That does not make this equipment obsolete, because it does exactly what is asked of it, but it offers none of the features modern security mechanisms rely on, and replacing it is usually neither desirable nor affordable. It is precisely from that starting point that, in our view, OT retrofit remains possible and deserves to be revisited using dedicated add-on components.</p>
<h2 id="modifying-the-machine-the-costs-rarely-anticipated">Modifying the machine: the costs rarely anticipated</h2>
<p>The temptation is to bring the machine &quot;up to standard&quot; by working directly on it, but touching the core of an industrial asset triggers a chain of consequences that tend to be underestimated at the moment of a purely financial decision (&quot;this must not cost more than buying a newer machine&quot;) or a purely technical one (&quot;if we can pull it off at all, that will do&quot;).</p>
<h3 id="compliance-called-into-question">Compliance called into question</h3>
<p>A machine carries a CE marking and a safety assessment that rest on a specific configuration, and opening the cabinet, rewiring, or inserting a module into the circuit can be enough to put that compliance in doubt, and may even trigger a fresh risk assessment.</p>
<h3 id="liability-shifted-onto-you">Liability shifted onto you</h3>
<p>Working on a manufacturer&#39;s equipment frequently means taking on part of their warranty, because what used to be their responsibility becomes, the moment the cover comes off, partly yours.</p>
<h3 id="availability-compromised">Availability compromised</h3>
<p>Any intervention on equipment in production means downtime, scheduling and risk, and on a critical site that is far from trivial.</p>
<h3 id="maintenance-debt">Maintenance debt</h3>
<p>Every modification eventually becomes a special case, one that has to be documented, maintained and understood for the entire service life of the machine, and taken together these constraints mean you often end up inheriting the very problem you set out to solve.</p>
<h2 id="a-module-inside-the-cabinet-moves-the-problem-it-does-not-so">A module inside the cabinet moves the problem, it does not solve it</h2>
<p>One variant looks smarter, since rather than reworking the machine, you add an intelligent module to it, but the moment you enter the equipment, you modify it, with exactly the same consequences as above.</p>
<p>In doing so, you also give up two forms of control. <strong>Cost</strong>, first, because every integration becomes bespoke work that is hard to replicate from one site to the next, and <strong>security</strong>, second, because you are grafting a component inside a system you never fully know, and in environments this demanding that is a lot of unknowns.</p>
<h2 id="our-choice-for-ot-retrofit-enforcement-at-the-boundary-of-th">Our choice for OT retrofit: enforcement at the boundary of the machine</h2>
<p>So we preferred another route, placing standalone units that sit <em>at the boundary</em> of the equipment, on the power cable, on the radio link, on the serial bus or on the network port, rather than inside it.</p>
<p>The model we follow fits into a single image, an anchor and satellites, because a decision unit, which we call the Cube, acts as the trust anchor for a zone, and effectors, the satellites, enforce the decision, one per interface.</p>
<p>The existing equipment itself is never touched, so its compliance remains intact, its manufacturer stays solely responsible for its core, and the whole arrangement can be removed without leaving a trace.</p>
<p>This is not an absolute rule, and other approaches have their merits, but it is a choice we stand behind, because it preserves what already works.</p>
<h2 id="keeping-custom-development-to-a-minimum">Keeping custom development to a minimum</h2>
<p>This choice would be of limited interest if it forced us to reinvent every building block, so we hold ourselves to a companion rule, which is to keep custom development to a strict minimum, assembling proven modules rather than designing bespoke ones.</p>
<h3 id="the-shelly-example-a-lot-of-function-in-a-very-small-box">The Shelly example: a lot of function in a very small box</h3>
<p>The <a href="https://www.shelly.com/">Shelly</a> module is a good illustration of why we prefer to assemble rather than design, because a tiny, inexpensive component already brings together Wi-Fi, Bluetooth Low Energy, dedicated protocols such as Matter, power metering and a switching relay.</p>
<p>The whole thing fits inside a sealed enclosure with two simple cable glands, one in and one out, which is enough to protect equipment outdoors or in a damp environment, and the ratio between price and built-in functionality is, quite honestly, hard to match with in-house development.</p>
<p>From this module, we keep only the behaviour we want, because commands go through our unit, the physical input merely signals, and the relay obeys nothing but the local decision, and we can disable every ancillary feature, in particular the vendor cloud and smartphone app control.</p>
<h3 id="the-security-layers-you-can-stack-on-top">The security layers you can stack on top</h3>
<p>This small module is not just versatile, because it already offers several layers of security to build on, and others that can be enabled or added on top, and we summarise them here since they deserve an article of their own.</p>
<p>At the network level, Wi-Fi with WPA2 / WPA3 and MQTT protected by mTLS encrypt and authenticate the exchanges, and Matter brings its own model, with certificate-based device attestation and secure commissioning. The stock firmware updates over a signed channel, which limits the risk of remote tampering, and local access (web interface, RPC calls) can be authenticated, with the physical input decoupled from the relay so that the button signals without ever directly commanding.</p>
<p>On top of all that, we add our own layer, with signed ephemeral capabilities, a decision taken locally, and a default-to-stop behaviour whenever there is doubt.</p>
<p>The point, at this stage, is simply that an off-the-shelf module can carry serious security without heavy custom development, and we will detail these protocols, and how to combine them, in a follow-up article.</p>
<h3 id="open-source-building-blocks-in-software-and-in-hardware">Open source building blocks, in software and in hardware</h3>
<p>For everything else, we favour open source building blocks, in both software and hardware.</p>
<p>The firmware we develop is released under a free licence and builds on open, audited frameworks such as Espressif&#39;s <a href="https://github.com/espressif/esp-idf">ESP-IDF</a> or the Linux Foundation&#39;s <a href="https://www.zephyrproject.org/">Zephyr RTOS</a>. On the hardware side, we build on open platforms, documented microcontroller boards, shared designs, open hardware cellular modules such as <a href="https://www.dptechnics.com/en/products/walter.html">Walter</a>, whose schematics and KiCad footprints are published, which let us understand exactly what we are deploying.</p>
<p>The work we keep in-house then concentrates on what actually creates our value, the chain of trust that ties these building blocks together, not the plumbing that makes them run.</p>
<h3 id="the-double-effect-cost-and-security">The double effect: cost and security</h3>
<p>This position has a double effect, and that is why we hold to it, because it controls <strong>cost</strong>, since an off-the-shelf module has a price known in advance and repeatable from one site to the next, and it also controls <strong>security</strong>, since a component we have chosen, whose firmware and behaviour we know, can be audited and hardened.</p>
<h2 id="interfacing-with-legacy-equipment-interface-by-interface">Interfacing with legacy equipment, interface by interface</h2>
<h3 id="one-pattern-applied-to-every-interface">One pattern, applied to every interface</h3>
<p>The other merit of the add-on approach is that it stays largely agnostic to the machine being protected, the same pattern playing out interface by interface: cutting a 230 V supply, filtering a USB keyboard or mouse, proxying a wireless peripheral, segmenting an Ethernet port, or sitting in line on an RS-485 / Modbus serial bus.</p>
<p>In each case, an ageing asset is enrolled without being modified, because you observe it, you surround it, and you act at its boundary.</p>
<h3 id="what-works-today-what-is-coming">What works today, what is coming</h3>
<p>To be clear, the effector that is fully operational today is power cut-off, and we have explained elsewhere why cutting the power is not an emergency stop, while the other interfaces are on our roadmap and rest on the same principle, with first look going to the customers who come to us with a new edge case.</p>
<h2 id="concrete-building-blocks-and-where-to-find-them">Concrete building blocks, and where to find them</h2>
<p>To make all this tangible, here are a few examples of OT retrofit without touching the machine, and the solutions, commercial or open source, to build on.</p>
<h3 id="four-things-you-can-do-without-opening-anything">Four things you can do without opening anything</h3>
<ul>
<li>Cutting power to an uncooperative variable-speed drive using a Shelly Pro 1PM clipped onto the DIN rail in the cabinet, without ever opening the drive itself.</li>
<li>Spotting a stray BLE tracker, or a suspicious 433 MHz remote, with a listening radio gateway.</li>
<li>Isolating a PLC segment behind a network gateway, without having to reconfigure the PLC.</li>
<li>Talking read-only to an ageing Modbus PLC to inventory it, without ever writing to it.</li>
</ul>
<h3 id="the-toolbox">The toolbox</h3>
<ul>
<li><strong>Switching and energy metering</strong>: <a href="https://www.shelly.com/">Shelly</a> (off-the-shelf modules, from DIN rail to plug), with <a href="https://esphome.io/">ESPHome</a> or <a href="https://tasmota.github.io/docs/">Tasmota</a> as free alternative firmware.</li>
<li><strong>Open source embedded</strong>: <a href="https://github.com/espressif/esp-idf">ESP-IDF</a> and <a href="https://www.zephyrproject.org/">Zephyr RTOS</a> for software, and <a href="https://www.dptechnics.com/en/products/walter.html">Walter</a> with its <a href="https://github.com/QuickSpot">QuickSpot</a> repositories for open cellular hardware.</li>
<li><strong>Hardware root of trust</strong>: STMicroelectronics&#39; <a href="https://www.st.com/en/secure-mcus/stsafe-a120.html">STSAFE-A120</a> secure element, on a Common Criteria EAL5+ certified platform, complemented by the GSMA&#39;s IoT SAFE framework.</li>
<li><strong>BLE and sub-GHz listening</strong>: <a href="https://github.com/1technophile/OpenMQTTGateway">OpenMQTTGateway</a> (BLE and 433 / 868 MHz) and the <a href="https://github.com/merbanan/rtl_433">rtl_433</a> decoders.</li>
<li><strong>Network and fieldbus</strong>: <a href="https://openwrt.org/">OpenWrt</a> to segment an Ethernet port, <a href="https://libmodbus.org/">libmodbus</a> to interface with an RS-485 / Modbus bus.</li>
<li><strong>Satellite design</strong>: the free electronics CAD tool <a href="https://www.kicad.org/">KiCad</a>, which we also use to publish our own schematics.</li>
</ul>
<p>None of these building blocks is a secret, and that is precisely the point, because we save the bespoke work for what really makes the difference, the chain of trust that ties them together.</p>
<h2 id="a-successful-ot-retrofit-is-a-layer-not-a-graft">A successful OT retrofit is a layer, not a graft</h2>
<p>Grafting an organ always means risking rejection, whereas adding a layer stays reversible, costable and under control.</p>
<p>We do not claim, in doing this, to make the machine &quot;new&quot;, because prolonged physical access remains physical access, and we would rather document that than deny it, as we explain in <a href="/blog/remote-access-and-physical-access-two-different-problems">our article on remote access and physical access</a>.</p>
<p>This approach has the merit of bringing trust back where there was none, without imposing anything on what already works, and to our minds it is the most honest way to carry out an OT retrofit on an installed base you cannot replace.</p>
<p><em>Want to see the principle in action? In a demonstration, you take the intruder&#39;s seat, you try to defeat the system, it responds in front of you, and you leave with the signed audit log of your own attempts.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>IAM at the End of the Power Cord, Part 3</title>
      <link>https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-3</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-3</guid>
      <pubDate>Thu, 20 Aug 2026 11:53:14 GMT</pubDate>
      
      <description>Part three of our series on IAM at the end of the power cord: what we would do if we led an IAM program, and your IAM vocabulary translated row by row.</description>
      <content:encoded><![CDATA[<p>This third part of our series on IAM at the end of the power cord moves from analysis to action, with what we would do if we led an IAM program and your vocabulary translated row by row.</p>
<p><em>This article follows <a href="/blog/iam-at-the-end-of-the-power-cord-part-1">part one</a> and <a href="/blog/iam-at-the-end-of-the-power-cord-part-2">part two</a> of the series.</em></p>
<h2 id="what-to-do-if-you-lead-an-identity-and-access-management-iam">What to do if you lead an identity and access management (IAM) program</h2>
<p><strong>Inventory critical assets lacking access policies.</strong> Not servers or applications, but machines, cabinets, test benches and equipment where unauthorized use carries physical, safety or regulatory impact, and for which you cannot produce an access list or an audit log. In many industrial organizations this list easily rivals the number of governed applications, and it was simply never compiled because it fell outside everyone&#39;s scope, too physical for IT and too digital for facility maintenance.</p>
<p><strong>Define denial semantics equipment by equipment with the safety teams.</strong> In our experience this is the hardest and most structuring deliverable, because it says, machine by machine, what a denial produces (a stop, a logged continuation or a reduced mode), it dictates the whole architecture, and it cannot be done by the security team alone.</p>
<p><strong>Require an enforcement point capable of offline decision-making.</strong> This criterion quickly filters out unsuitable options, because any solution relying on a continuous link to a central service will fail during the first outage, which is precisely the moment when risk is highest.</p>
<p>And one thing not to do: <strong>do not connect equipment to the network solely in order to govern it</strong>, because physical access control does not require the machine to be smart, it requires the power outlet to be.</p>
<h2 id="your-vocabulary-translated">Your vocabulary translated</h2>
<table>
<thead>
<tr>
<th>IAM concept</th>
<th>On a sensitive machine</th>
<th>What breaks</th>
</tr>
</thead>
<tbody><tr>
<td><strong>Subject identity</strong></td>
<td>The person intervening: operator, in-house maintainer, tier-2 subcontractor, OEM field engineer</td>
<td>Usually <strong>not</strong> in the operator&#39;s directory. Federation with no shared IdP, and often no IdP at all on the supplier side.</td>
</tr>
<tr>
<td><strong>Resource identity</strong></td>
<td>The machine, the cabinet, the test bench</td>
<td><strong>No cryptographic identifier.</strong> IEEE 802.1AR-2018 (IDevID/LDevID) is the right normative answer, but it assumes an identity minted at the factory, whereas the installed base has a nameplate.</td>
</tr>
<tr>
<td><strong>Authentication</strong></td>
<td>Proof that the right person is physically present, here, now</td>
<td>No keyboard, no screen, no browser, so the factor must be carried by the person and validated by a reader external to the machine.</td>
</tr>
<tr>
<td><strong>Authorization</strong></td>
<td>Role <strong>and</strong> valid qualification <strong>and</strong> scheduled work order <strong>and</strong> zone</td>
<td>De facto ABAC, with one OT specificity, since <strong>purpose</strong> is a first-class attribute: being qualified is not enough, because you need a legitimate reason to be here <em>now</em>. NERC CIP-004-7 R4.1 puts electronic access and unescorted physical access under a single authorization process.</td>
</tr>
<tr>
<td><strong>PEP</strong></td>
<td>The actuator that cuts or releases power</td>
<td>Must be physically separate from the machine, or the machine judges its own case.</td>
</tr>
<tr>
<td><strong>PDP</strong></td>
<td>The policy engine</td>
<td><strong>Must be local.</strong> NIST SP 800-207 describes the PEP as <em>&quot;responsible for enabling, monitoring, and eventually terminating connections between a subject and an enterprise resource&quot;</em>, which assumes the connection exists, whereas here it does not.</td>
</tr>
<tr>
<td><strong>Session</strong></td>
<td>The period the machine is energized</td>
<td>Barely exists, because an energized machine stays energized, so it needs continuous re-evaluation, the physical counterpart of CAEP / Shared Signals.</td>
</tr>
<tr>
<td><strong>Provisioning / JML</strong></td>
<td>Joiner, mover, leaver, but mostly <strong>credential expiry</strong></td>
<td>The real OT leaver is a lapsed qualification, not an HR departure. NFPA 70E-2024 §110.4(A)(3) requires requalification at intervals not exceeding three years, and Australia&#39;s CIRMP rules require re-vetting of critical workers at least every five years. Note also the distinction NFPA 70E draws in §110.4(A)(1)(b), <em>&quot;a person shall be qualified for certain equipment and tasks to be performed&quot;</em>, which is least privilege written into an electrical safety standard.</td>
</tr>
<tr>
<td><strong>Access review</strong></td>
<td>Recertifying &quot;may energize machine X&quot;</td>
<td>NERC CIP-004-7 R4.2 already mandates verification of unescorted physical access authorizations <strong>once each calendar quarter</strong>, and outside that regime it is close to nonexistent.</td>
</tr>
<tr>
<td><strong>Logging</strong></td>
<td>Timestamped, attributed proof of use</td>
<td>NERC CIP-006-6 R1.8: log <strong>each individual</strong>, with information identifying them and the date and time of entry, and R1.9: retain <strong>90 calendar days</strong>. Elsewhere, the only log is the machine&#39;s hour counter, which is unattributed, unsigned and reset at maintenance.</td>
</tr>
<tr>
<td><strong>PAM</strong></td>
<td>The OEM engineer&#39;s privileged access</td>
<td>PAM covers the <strong>remote</strong> session, while the same engineer, with the same laptop and the same reprogramming capability, standing in front of the cabinet, is covered by nothing.</td>
</tr>
<tr>
<td><strong>Revocation</strong></td>
<td>Removing the right to use</td>
<td>NERC CIP-004-7 R5.1: unescorted physical access removed within <strong>24 hours</strong> of a termination action, against 30 calendar days for named logical accounts under R5.3. And here, revoking means <strong>cutting power</strong>, with physical consequences.</td>
</tr>
</tbody></table>
<p>Read that table honestly and you will notice that you already know how to handle ten rows out of twelve, at least conceptually, so what is missing is not IAM doctrine but an enforcement point that acts in the physical world and a decision point that survives OT conditions.</p>
<h2 id="expanding-iam-to-the-end-of-the-power-cord-not-the-stack">Expanding IAM to the end of the power cord, not the stack</h2>
<p>IAM spent fifteen years expanding its reach, from employees to customers, customers to partners, humans to workloads, workloads to autonomous agents, and at every step the same objection resurfaced, <em>&quot;that&#39;s not really identity&quot;</em>, and at every step it was proven wrong.</p>
<p>The next frontier belongs to assets that have no IP address and never will. They form a large part of the production core of European critical infrastructure, they are subject to the same regulatory pressures as everything else, and they are currently managed with physical keyrings and paper logbooks.</p>
<p>The good news is that your mental model holds up, identity, authentication, policy, enforcement, proof, review, and you simply need to accept that the enforcement point is no longer a software component on a network path but a hardware device on an electrical circuit, and that decisions must be made locally, even when connectivity completely drops.</p>
<p>This reasoning gave birth to Trust-Buster, with a local policy engine that establishes the identity and legitimacy of an intervention from its physical and radio-frequency environment, actuators that enforce decisions on the equipment&#39;s power supply, and an audit-ready proof log, all of it in nominal operation whether connected or offline.</p>
<p>IAM at the end of the power cord, in short, is the same discipline you practice, with a circuit breaker at the end of the chain.</p>
]]></content:encoded>
    </item>
    <item>
      <title>IAM at the End of the Power Cord, Part 2</title>
      <link>https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-2</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-2</guid>
      <pubDate>Thu, 20 Aug 2026 11:51:28 GMT</pubDate>
      
      <description>Part two of our series on IAM at the end of the power cord, where five IT reflexes become wrong in OT and where the regulations all end up asking for proof.</description>
      <content:encoded><![CDATA[<p>This second part of our series on IAM at the end of the power cord is where applying IT concepts to our domain gets tricky, because five principles you would normally apply without a second thought become wrong or dangerous here, and we go through them before rereading what the regulations already ask.</p>
<p><em>This article follows <a href="/blog/iam-at-the-end-of-the-power-cord-part-1">part one</a> of the series and continues in <a href="/blog/iam-at-the-end-of-the-power-cord-part-3">part three</a>.</em></p>
<h2 id="five-iam-reflexes-you-need-to-unlearn">Five IAM reflexes you need to unlearn</h2>
<blockquote>
<p>&quot;physical access control (guards, gates and guns) to protect a control room to restrict access to a group of known personnel <strong>to compensate for the technical requirement for personnel to be uniquely identified by the IACS</strong>&quot; (IEC 62443-4-2, definition 3.1.10, <em>compensating countermeasure</em>)</p>
</blockquote>
<p>The standards committee has in a sense written our thesis for us, because physical access control is what you deploy when the industrial system cannot tell one human being from another, and on the installed base we meet that is nearly always the case.</p>
<h3 id="1-when-in-doubt-deny-is-no-longer-a-universal-rule">1. &quot;When in doubt, deny&quot; is no longer a universal rule</h3>
<p>In IT, fail-closed is the default stance and trade-offs are almost always decided in favor of security, but in OT denying access can put people in danger, because cutting power to a ventilation system, a cooling unit, a safety loop or a piece of equipment in the middle of a critical cycle is a safety-of-life decision before it is a cybersecurity decision.</p>
<p>The practical consequence is that the semantics of denial must be defined equipment by equipment, and this decision does not belong to the security team alone, since it belongs jointly to security, operations and HSE (health, safety, environment). Some equipment will be <em>fail-secure</em> (power cut), other equipment <em>fail-safe</em> (power maintained, logged, alerted), and other equipment <em>fail-degraded</em> (reduced functionality allowed), so an OT policy model that cannot express these three outcomes seems unusable to us.</p>
<p>This is a nuance that canonical Zero Trust in IT has never had to address.</p>
<h3 id="2-connectivity-is-not-an-assumption">2. Connectivity is not an assumption</h3>
<p>You design your architectures assuming the identity provider (IdP) is reachable, but in industrial or defense environments that assumption collapses, because there are sites air-gapped by design, WAN outages lasting days, shielded facilities and contexts where connectivity is actively unwanted.</p>
<p>A policy decision point (the PDP of the IAM chain) that calls a remote service in order to decide is one that refuses to decide when it matters most, so the decision must be taken locally and offline, and it must remain nominal rather than degraded.</p>
<p>You already know this compromise in another form, the local verification of a signed token without calling back the issuer for introspection, and it raises the same questions with the same answers, namely signed and versioned policies, bounded lifetimes, revocation lists with an explicit maximum freshness, trusted time sources and reconciliation upon reconnection. What changes here is the stakes, because a stale token accepted by mistake does not merely grant access to an API, it grants access to a machine capable of physically hurting someone.</p>
<h3 id="3-equipment-identity-cannot-be-declarative">3. Equipment identity cannot be declarative</h3>
<p>In IT, we trust what the resource claims about itself because an agent attests to it, whereas in OT the equipment does not know who it is and cannot attest to anything, and worse, it is entirely replaceable, since nothing stops someone from unplugging the authorized machine and plugging another one into its place.</p>
<p>Hence an unintuitive but fruitful shift, because you no longer identify the equipment itself but the power point, and you continuously characterize what is connected to it. The electrical and radio-frequency signature of a plugged-in device is an observable attribute, and changes to it become identity events in their own right.</p>
<p>This creates a primitive that does not exist in traditional IAM, rewiring detection as a security signal, because a legitimate operator with valid entitlements and a compliant work order who plugs an unexpected piece of equipment into a controlled outlet produces an event that your policy must know how to qualify.</p>
<h3 id="4-the-proof-must-outlive-the-auditor-not-the-siem">4. The proof must outlive the auditor, not the SIEM</h3>
<p>Your logging is designed for detection, with high volume, low unit value, short retention and a network path to a central collector, the SIEM.</p>
<p>Here it is the exact opposite, because volume is low, a few dozen events per machine per month, unit value is extremely high, and the primary use case is almost never real-time detection but retrospective proof, examined eighteen months later by an insurer, a prime contractor, a regulator or an investigating judge to answer one simple question: <em>who powered on this machine on March 14 at 10:40 PM, and on what grounds?</em></p>
<p>A log sent over the network to a SIEM fails this requirement because the network was likely down on March 14, so you need a local, timestamped, chained, signed and non-repudiable log that can be exported independently of any central infrastructure, and we have written separately about what such a proof log proves and what it does not.</p>
<h3 id="5-the-subject-is-not-necessarily-human-nor-necessarily-yours">5. The subject is not necessarily human, nor necessarily yours</h3>
<p>You are already navigating the wave of non-human identities, service accounts, workloads, proliferating secrets and autonomous agents, but the OT world adds a layer IT rarely encounters, because a large share of legitimate subjects, often the majority on a given site, do not belong to the organization: temporary workers, tier-2 subcontractors, OEM technicians traveling from another continent, maintenance providers dispatched by a third party.</p>
<p>You must therefore federate an identity, verify an entitlement attribute and link it to an intervention order, without ever having provisioned the person, often without a directory on the vendor&#39;s side, and with no network connection at the time of decision. This is not a problem solved elsewhere and copied here, because it is still an open IAM problem, and one we find fascinating.</p>
<h2 id="what-the-regulations-already-say-that-few-have-read">What the regulations already say (that few have read)</h2>
<p>The idea that IAM belongs to IT while safety belongs to OT does not survive contact with the standards.</p>
<h3 id="what-every-standards-body-already-agrees-on">What every standards body already agrees on</h3>
<p><strong>IEC 62443</strong> is built on seven foundational requirements, and the first one, FR 1, is <em>Identification and Authentication Control</em>, which explicitly distinguishes three classes of subject: human user, software process and device. The reference security standard for industrial automation puts IAM at the top of its list, and has for years, so rather than an IT import it is a chapter that industry never managed to apply to the installed base, for lack of a technical means to do so.</p>
<p>The same family goes further than most people realise, because IEC 62443-4-2 requires embedded devices to protect their physical diagnostic and test interfaces against unauthorised use (CR/EDR 2.13) and to provide physical tamper resistance and detection (CR/EDR 3.11), mandatory from security level 2, with logging and notification of tampering attempts from level 3.</p>
<p>It also states the honest objection, and we should state it too, since identification and authentication <em>&quot;should not hamper fast, local emergency actions.&quot;</em> That constraint is real, and it is exactly why the meaning of &quot;deny&quot; has to be defined machine by machine.</p>
<p><strong>NIST SP 800-82r3</strong>, the <em>Guide to Operational Technology Security</em>, places physical security as layer 2 of 5 in its defence-in-depth model, before the network, and its OT tailoring of control PE-3 is unambiguous, because physical access controls <em>&quot;are used by the organization when necessary and possible to supplement OT security when electronic mechanisms are unable to fulfill the security requirements of the organization&#39;s security plan.&quot;</em></p>
<p><strong>ISO/IEC 27001:2022</strong> restructured Annex A around four themes, one of which is Physical controls (7.1 to 7.14), and two matter here: 7.11 <em>Supporting utilities</em>, the only control in the annex that treats electrical power as a security asset in its own right, and 7.4 <em>Physical security monitoring</em>, new in the 2022 revision and the physical counterpart of logging. The detail sits in ISO/IEC 27002:2022, clause 7.</p>
<h3 id="where-it-is-already-legally-binding">Where it is already legally binding</h3>
<table>
<thead>
<tr>
<th>Jurisdiction</th>
<th>Instrument</th>
<th>What it demands</th>
</tr>
</thead>
<tbody><tr>
<td><strong>North America</strong> (electricity)</td>
<td>NERC CIP, enforced by FERC</td>
<td>Authorization of physical and electronic access under one process (CIP-004-7 R4.1), quarterly verification (R4.2), removal of unescorted physical access within 24 hours (R5.1), two or more physical access controls where technically feasible (CIP-006-6 R1.3), alarm within 15 minutes of detection (R1.5), per-individual entry logging (R1.8), 90-day retention (R1.9), escorted visitors with a named point of contact (R2.2), and testing of the access control system every 24 months (R3.1)</td>
</tr>
<tr>
<td><strong>European Union</strong></td>
<td>NIS2 (2022/2555) and CER (2022/2557)</td>
<td>NIS2 Art. 21(2): access control policies, asset management, HR security, multi-factor authentication, supplier security. CER Art. 13, which most commentary misses, requires <em>&quot;adequate physical protection&quot;</em> including access controls, plus employee security management establishing access rights to premises, critical infrastructure and sensitive information by category of critical personnel, and Art. 14 adds background checks.</td>
</tr>
<tr>
<td><strong>European Union</strong></td>
<td>Cyber Resilience Act (2024/2847)</td>
<td>Annex I requires products with digital elements to be protected from unauthorised access <em>&quot;by appropriate control mechanisms, including but not limited to authentication, identity or access management systems&quot;</em>, and to report on possible unauthorised access. It applies to any manufacturer placing a product on the Union market, wherever they are established, so a US, Japanese or Taiwanese vendor selling into the EU is in scope and must have an authorised representative or importer carrying the obligations.</td>
</tr>
<tr>
<td><strong>Australia</strong></td>
<td>Security of Critical Infrastructure Act 2018 and the CIRMP Rules</td>
<td>The only regime we know of that names both vectors in the rules themselves, with a <em>Personnel hazards</em> rule (background checks, periodic re-vetting, plans to minimise unauthorised, unescorted or privileged access to critical components) and a <em>Physical security hazards</em> rule (documented processes to control and limit physical access to critical components, centralised management, continuous monitoring of access controls). A recent tightening moved the regime from principles-based to substantially more prescriptive.</td>
</tr>
<tr>
<td><strong>Germany</strong></td>
<td>Critical infrastructure umbrella act (KRITIS-Dachgesetz), implementing CER</td>
<td>Perimeter security, personnel and vehicle access control, staff vetting and training, written resilience plans, 24-hour incident notification. Outcome-based rather than prescriptive, since the operator determines what is appropriate and proportionate, and has to document it.</td>
</tr>
<tr>
<td><strong>Singapore</strong></td>
<td>Cybersecurity Act and the Code of Practice for CII (CCoP2.0)</td>
<td>Carries an explicit OT addendum, acknowledging that standard IT tooling does not speak Modbus, DNP3 or IEC 61850, and requires detection of lateral movement and credential misuse across both IT and OT, with auditable log retention available on demand for investigation.</td>
</tr>
</tbody></table>
<h3 id="what-is-coming">What is coming</h3>
<p>The United Kingdom is not covered by NIS2, and its Cyber Security and Resilience Bill, still a Bill in the House of Lords at the time of writing, will amend the NIS Regulations 2018 and pull managed service providers and designated critical suppliers into scope.</p>
<p>In the United States, the picture is less flattering than the European one and worth stating plainly, because CIRCIA was enacted in 2022 and its final rule has slipped well past its statutory deadline, and the TSA&#39;s 2024 proposed rule for surface transport has not been finalised, so outside NERC&#39;s electricity sector the binding American obligations are thinner than the headlines suggest.</p>
<p>Timelines aside, the pressure that actually reaches a plant manager rarely comes from a legislature, because it comes down the supply chain, in the security questionnaires and cyber clauses that automotive, aerospace, energy and defence primes are already sending to their tier-2 and tier-3 suppliers, and we have written about <a href="/blog/how-to-navigate-nis2-compliance-for-your-business">where to start with NIS2</a> and about <a href="/blog/emerging-trends-in-cybersecurity-legislation-what-to-expect-in-2027">what to expect in 2027</a>.</p>
<h3 id="what-they-all-demand-of-iam-at-the-end-of-the-power-cord-evi">What they all demand of IAM at the end of the power cord: evidence</h3>
<p>These regimes differ in scope, in force and in style, but they converge on one thing, and it is the blind spot, because they all demand proof, proof of access control, proof of traceability, proof that qualifications were current, and proof of <em>physical</em> access to an industrial machine exists, as far as we can tell, almost nowhere. That is the gap from which <a href="/blog/iam-at-the-end-of-the-power-cord-part-3">part three</a> of this series on IAM at the end of the power cord starts.</p>
]]></content:encoded>
    </item>
    <item>
      <title>IAM at the End of the Power Cord, Part 1</title>
      <link>https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-1</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/iam-at-the-end-of-the-power-cord-part-1</guid>
      <pubDate>Thu, 20 Aug 2026 11:45:52 GMT</pubDate>
      
      <description>Why your last policy enforcement point is a power outlet, and why IAM at the end of the power cord is where identity and access management is still missing.</description>
      <content:encoded><![CDATA[<p class="lead"><strong>Why your last policy enforcement point is a power outlet</strong></p>
<p><em>This series on IAM at the end of the power cord is intended for IAM practitioners, and you will find here the transposition of identity and access management into a domain where it is essential and, to this day, almost completely absent.</em></p>
<p>In North America, a grid operator must remove a departing employee&#39;s unescorted physical access to a substation within 24 hours of the termination action, whereas that employee&#39;s named user accounts can wait thirty days.</p>
<p>That is not an oversight, because it is NERC CIP-004-7, R5.1 and R5.3, and it is legally enforceable with penalties that run per day and per violation, so the North American regulator has decided that the physical door is thirty times more urgent to close than the login.</p>
<p>Now ask yourself which of the two your IAM programme actually governs.</p>
<h2 id="where-your-iam-chain-physically-ends">Where your IAM chain physically ends</h2>
<p>Let&#39;s review the chain using the vocabulary we all share, because an identity is established and provisioned, an authenticator proves its possession, a PDP (the policy decision point) evaluates a policy, a PEP (the policy enforcement point) enforces the decision along the access path, a session is opened then re-evaluated, a log generates proof, and a review closes the loop.</p>
<p>This chain relies on an assumption we no longer even formulate because it feels so obvious, which is that there is a digital path between the subject and the resource and that we can insert ourselves into it, through a reverse proxy, an agent, a SCIM connector, a PAM broker, a domain controller or an application gateway, since the PEP is a piece of software sitting on this path.</p>
<p>Remove the digital path, and the entire chain becomes inapplicable, yet this is the reality for the overwhelming majority of sensitive industrial equipment in service today.</p>
<ul>
<li>A PLC installed in 2003 whose only access control is a key switch on the front panel, with three keys in circulation and a fourth lost six years ago.</li>
<li>An electrical cabinet whose opening is protected by a triangular key you can buy online for the price of a coffee.</li>
<li>A test bench, a furnace, an autoclave, a press, an EDM machine, in other words equipment where unauthorised use is not a confidentiality risk but a risk of physical safety, counterfeiting, sabotage or misuse.</li>
<li>Equipment connected to the site&#39;s power grid, but deliberately disconnected from the IT network, precisely because its operator judged, often correctly, that the connection would create more risk than it would mitigate.</li>
</ul>
<p>Faced with this asset base, the knee-jerk reaction of the IT world is to demand network connectivity, <em>&quot;connect it, and we will govern it&quot;</em>, but this is an expensive, slow and often counterproductive response, because connecting a machine, and especially every machine, sometimes designed at a time when network security was not part of any specification, just to monitor it, amounts to creating an attack surface simply to monitor the attack surface you have just created.</p>
<p>There is another way.</p>
<h2 id="our-thesis-iam-ends-at-the-power-cord-the-only-universal-enf">Our thesis: IAM ends at the power cord, the only universal enforcement point</h2>
<p>All of this equipment shares one common trait, without exception, because it is powered by electricity.</p>
<p>A machine without power does not produce, does not heat up, does not rotate, does not communicate and cannot be reprogrammed, as long as it has no energy reserve of its own, so the power supply is the prerequisite for almost any use, and from an IAM architect&#39;s perspective it possesses remarkable properties.</p>
<ul>
<li><strong>It is universal.</strong> No assumptions about protocol, firmware, age, manufacturer or the presence of an IP stack, and the control applies identically to a 1998 PLC and to a machine delivered last month.</li>
<li><strong>It precedes the logical layer.</strong> A software PEP executes after the target boots and depends on its integrity, whereas an electrical PEP applies before the target exists functionally, so no firmware vulnerability on the equipment allows bypassing the absence of power. It is a <em>deny</em> that the resource itself cannot bypass, a rare property we almost never manage to achieve in software architectures.</li>
<li><strong>It is binary and stateless.</strong> No session to negotiate, no context to maintain in the equipment, no client to install.</li>
</ul>
<p>The corollary is straightforward, because in a sensitive installation the power outlet is, in our view, the enforcement point of first and last resort, with one reserve we detail elsewhere, since power off is not an emergency stop.</p>
<p>This idea is not new, it is simply ungoverned, and three precedents should reassure any IAM practitioner that we are not proposing something exotic.</p>
<p><strong>Locking out a circuit.</strong> Across Europe, EN 50110-1 sets out five safety rules to be performed in order before work begins, <em>disconnect completely, secure against re-connection, verify absence of operating voltage, carry out earthing and short-circuiting, provide protection against adjacent live parts</em>, and Germany implements it as DIN VDE 0105-100 while the UK adopts it as BS EN 50110-1 on top of the Electricity at Work Regulations 1989.</p>
<p>Its 2023 edition went further, and in a direction that should interest you, because it defines a role taxonomy (installation manager, operation controller, work controller, worker) and separates authorization (&quot;formal approval to perform planned work&quot;) from permission to start work (&quot;direct instruction to the workers at work location to commence work after all safety measures are taken&quot;), which is a static entitlement granted upstream and a dynamic, contextual grant issued on site, that is a PDP and a PEP described in an electrical safety standard, in 2023.</p>
<p>In the United States, OSHA 29 CFR 1910.147, the lockout/tagout standard, makes the point even more bluntly, because on lockout devices §(c)(5)(ii)(D) requires that they <em>&quot;shall indicate the identity of the employee applying the device(s)&quot;</em>, so a padlock that carries a name is a credential, and a group lockbox with one personal lock per authorized employee is a multi-holder authorization scheme with non-delegable release, since §(e)(3) requires the device be removed by the employee who applied it.</p>
<p>So the industrial world did invent electrical IAM, but an IAM with no directory, no automatic expiry, no queryable log and no remote revocation, which is 1970s IAM, and it is still how most plants run today.</p>
<p><strong>Switched PDUs.</strong> In any data centre, per-outlet authorization has existed for a quarter of a century, since you can cut and restore a power port remotely, but the subject of that authorization is an administrative account, never a business identity carrying a qualification, so it is an actuator with no policy.</p>
<p><strong>ISO 15118 and Plug &amp; Charge.</strong> The most interesting precedent, and the one every IAM practitioner should know, because for EV charging the industry specified X.509 certificate-based authentication of both endpoints, exchanged over the charging cable itself, by power-line communication on the pilot line, and anchored in a dedicated trust hierarchy, with a V2G root, contract certificates issued by the mobility operator, charge point certificates and OEM provisioning certificates minted at the factory. The first edition (ISO 15118-2) authenticated the charge point at the transport layer and the vehicle at the application layer, while ISO 15118-20 now mandates mutual TLS, so this is a full PKI, with a full certificate lifecycle, whose only purpose is to decide whether electricity should flow, and who gets billed.</p>
<p>The automotive world has already accepted that a power cable is an identity channel, while the industrial world is still on padlocks. That is where IAM at the end of the power cord begins, and <a href="/blog/iam-at-the-end-of-the-power-cord-part-2">Part 2</a> covers the five IAM reflexes you will have to unlearn before applying it.</p>
]]></content:encoded>
    </item>
    <item>
      <title>Emerging Trends in Cybersecurity Legislation: What to Expect in 2027</title>
      <link>https://trust-buster.com/blog/emerging-trends-in-cybersecurity-legislation-what-to-expect-in-2027</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/emerging-trends-in-cybersecurity-legislation-what-to-expect-in-2027</guid>
      <pubDate>Wed, 19 Aug 2026 11:54:07 GMT</pubDate>
      
      <description>Two EU regulations start applying in 2027 with no transition window, and the 2027 cybersecurity legislation changes what a CE mark on a machine certifies.</description>
      <content:encoded><![CDATA[<p class="lead"><strong>Two hard deadlines, no transition period, and one clause almost nobody has read</strong></p>
<p><em>The landscape of cybersecurity legislation is no longer evolving, because in 2027 it lands, as two EU regulations start applying within eleven months of each other, both without a transition window, and between them they change what a CE mark on an industrial machine actually certifies. That pair of texts is what we call the 2027 cybersecurity legislation in this article.</em></p>
<p><em>If you operate machinery in Europe, or sell it there, 2027 is not a year to watch, but a year to have prepared for.</em></p>
<h2 id="the-2027-cybersecurity-legislation-is-a-matter-of-deadlines-">The 2027 cybersecurity legislation is a matter of deadlines, not trends</h2>
<p>Most legislative round-ups hedge, but this one does not need to, because the two dates that matter in the 2027 cybersecurity legislation are already fixed in published regulations and neither has a grace period.</p>
<table>
<thead>
<tr>
<th>Date</th>
<th>Instrument</th>
<th>What changes</th>
</tr>
</thead>
<tbody><tr>
<td><strong>20 January 2027</strong></td>
<td>Machinery Regulation (EU) 2023/1230</td>
<td>Replaces the Machinery Directive outright, and cybersecurity becomes an essential health and safety requirement for CE marking.</td>
</tr>
<tr>
<td><strong>11 December 2027</strong></td>
<td>Cyber Resilience Act (EU) 2024/2847</td>
<td>Full application, and every product with digital elements placed on the Union market carries security obligations across its lifecycle.</td>
</tr>
</tbody></table>
<p>The absence of a transition window is the part that catches people, because a machine placed on the market on 19 January 2027 is governed by the old Directive while the same machine placed on the market a day later is governed by the Regulation, and there is no mixed regime and no opt-in period. For a reader coming to this cold, the CE mark is the manufacturer&#39;s declaration that a machine meets the EU&#39;s essential requirements, so changing what that mark certifies changes what a manufacturer has to prove before selling.</p>
<h2 id="20-january-2027-cybersecurity-becomes-a-safety-requirement">20 January 2027: cybersecurity becomes a safety requirement</h2>
<p>The Machinery Regulation replaces Directive 2006/42/EC, the framework that has governed the CE marking of machinery since the 2000s, and most of the coverage focuses on machine learning and notified bodies, which are the independent laboratories designated by member states to assess conformity, while for anyone working on OT security the more consequential change is quieter.</p>
<p>Cybersecurity moves into the essential health and safety requirements, since Annex III introduces a section titled <em>protection against corruption</em> (§1.1.9), alongside reinforced requirements on the safety and reliability of control systems (§1.2.1). In substance, connection to other systems or remote access must not create a hazardous situation, and safety-related software, data and hardware components must be identified as such and protected against corruption, whether accidental or deliberate.</p>
<p>That is a genuine shift in legal character, because under the Directive protecting a machine against tampering was a matter for the manufacturer&#39;s judgement, whereas from January 2027 it is a condition of placing the machine on the market at all. Industry associations asked for the cybersecurity requirements to be deferred to the CRA date, but as far as we know that deferral has not been granted.</p>
<h3 id="the-clause-almost-nobody-has-read-substantial-modification">The clause almost nobody has read: substantial modification</h3>
<p>Here is the provision that ought to be on the desk of every maintenance and engineering manager in Europe, and rarely is.</p>
<p>The Machinery Regulation defines substantial modification in EU law for the first time, and the consequence is short and severe, since whoever performs a substantial modification becomes the manufacturer for the aspects affected, with the attendant conformity assessment, technical documentation and liability.</p>
<p>Read that alongside how OT security retrofits have typically been done, because you open the cabinet, insert a module into the safety-related control circuit, rewire and close the cabinet. In 2026 that is a warranty conversation and a risk-assessment question, but from 2027 it is a question about whether your maintenance team has just become a machinery manufacturer, in a domain where the requirements now include cybersecurity of safety functions.</p>
<p>We have argued elsewhere that <a href="/blog/ot-retrofit-addons-instead-of-upgrades">OT retrofits should be add-ons rather than upgrades</a> and should sit at the boundary of a machine rather than inside it, on cost and liability grounds. The Machinery Regulation does not make that argument for us, but it does something more useful, which is to write the liability half of it into directly applicable EU law. Any 2027 retrofit plan should therefore have an explicit answer to the question <em>does this constitute a substantial modification?</em>, and the safest answer is the one where nothing inside the machine was touched.</p>
<h2 id="11-december-2027-the-cra-reaches-full-application">11 December 2027: the CRA reaches full application</h2>
<p>The Cyber Resilience Act entered into force on 10 December 2024, and its reporting obligations, for actively exploited vulnerabilities and severe incidents, apply from 11 September 2026. Full application arrives on 11 December 2027, and it is broad, since it covers the essential requirements in Annex I, technical documentation, vulnerability handling across a defined support period, and a conformity route that varies with product class.</p>
<p>Two points matter more than the headlines.</p>
<p>It is extraterritorial, because the CRA binds any manufacturer placing a product with digital elements on the Union market, wherever they are established, so a US, Japanese or Taiwanese vendor selling into the EU is in scope and must have an authorised representative or importer carrying the obligations. Readers outside Europe who have filed the CRA under &quot;European problem&quot; have filed it wrongly.</p>
<p>Machinery manufacturers are in both regimes at once, since the CRA&#39;s own recitals make clear that machinery makers must comply with both instruments, and the Commission and the European standardisation organisations are expected to promote consistency in how cybersecurity risk is assessed across the two, but that consistency is a work in progress rather than a settled fact, and we come back to it below.</p>
<p>And on access control specifically, Annex I is unusually direct for a product regulation, because products must be protected from unauthorised access by appropriate control mechanisms, <em>including but not limited to authentication, identity or access management systems</em>, and must report on possible unauthorised access. That is a product regulation using the phrase &quot;identity or access management systems&quot; in its operative text.</p>
<h2 id="standardisation-what-to-expect-and-when">Standardisation: what to expect, and when</h2>
<p>The Cyber Resilience Act does introduce a unified framework for cybersecurity requirements, and standardisation is genuinely the trend to watch, but it is worth being honest about where that standardisation currently stands rather than describing it as accomplished.</p>
<p>Harmonised standards are what turn an essential requirement into a testable specification and give manufacturers a presumption of conformity, and for machinery cybersecurity that layer is still being built, because EN 50742, addressing protection against corruption in machinery, is under development, and sector work such as the 2025 revision of ISO 10218-2 for industrial robots has begun folding cybersecurity requirements into safety standards. For the CRA, the standardisation request is substantial and the timetable is tight against December 2027.</p>
<p>The practical consequence for a manufacturer is uncomfortable but manageable, because you may reach an application date before the harmonised standard you would have relied on is cited in the Official Journal, and that does not suspend the requirement. It means demonstrating conformity by other means, namely a documented risk assessment, recognised international standards such as IEC 62443, and evidence you can put in front of an assessor.</p>
<p>The practical consequence for an operator is different and more interesting, because machines placed on the market from 2027 will arrive with tamper protection, software identity and logging that today&#39;s installed base simply does not have. Your estate becomes two-speed, with a growing minority of assets with usable security primitives and a large majority (the presses, drives, benches and cabinets commissioned between 1995 and 2026) that will never have them and cannot be retrofitted internally without the substantial-modification question above.</p>
<p>Governing that majority is not a 2027 problem, it is the problem the 2027 cybersecurity legislation makes visible.</p>
<h2 id="also-landing-and-worth-tracking">Also landing, and worth tracking</h2>
<ul>
<li><strong>AI Act, 2 August 2028.</strong> <a href="https://eur-lex.europa.eu/eli/reg/2026/1744/oj">Regulation (EU) 2026/1744 of 8 July 2026</a>, known as the Digital Omnibus on AI, postponed to 2 August 2028 the obligations for high-risk AI systems linked to Annex I products, and to 2 December 2027 those for Annex III systems. It also moved the Machinery Regulation to Section B of Annex I of the AI Act. For a machine, AI-specific requirements therefore go through delegated acts supplementing the Machinery Regulation, applicable from 2 August 2028. If you embed a machine-learning safety component, have it reviewed by a specialist.</li>
<li><strong>NIS2 and CER enforcement.</strong> The transposition phase is ending and the supervision phase is beginning, because the Commission has moved from infringement proceedings to referrals before the Court of Justice, and national authorities that spent 2025 and 2026 registering entities will spend 2027 inspecting them. We have written separately about <a href="/blog/how-to-navigate-nis2-compliance-for-your-business">how to navigate NIS2 compliance</a>.</li>
<li><strong>United Kingdom.</strong> The Cyber Security and Resilience Bill remains a Bill, in the House of Lords at the time of writing, and it will amend the NIS Regulations 2018 and bring managed service providers and designated critical suppliers into scope.</li>
<li><strong>United States.</strong> The picture is less flattering than the European one, and worth stating plainly, because the final rule under <a href="https://www.federalregister.gov/documents/2024/04/04/2024-06526/cyber-incident-reporting-for-critical-infrastructure-act-circia-reporting-requirements">CIRCIA</a> has slipped well past its statutory deadline, and the TSA&#39;s 2024 proposed rule for surface transport is not finalised. Outside NERC CIP in the electricity sector, binding American obligations remain thinner than the headlines suggest.</li>
</ul>
<h2 id="what-to-do-with-the-time-left">What to do with the time left</h2>
<ol>
<li><strong>Map your product range or your estate against the 20 January 2027 cut.</strong> For manufacturers, which products change Annex I category, and which lose the self-certification route, and for operators, which machines you expect to purchase after that date, and what security primitives they will arrive with.</li>
<li><strong>Put the substantial-modification question into your retrofit approval process</strong>, before the engineering and not after, because it is a one-line addition that can prevent an expensive category error.</li>
<li><strong>Review how software identity and tampering evidence will be captured and stored</strong>, since both regimes ask for it, and neither accepts &quot;the machine&#39;s hour counter&quot; as an answer.</li>
<li><strong>Compile the OT asset inventory</strong> if you have not already, because every obligation above is scoped by a list most organisations have never produced.</li>
<li><strong>Do not wait for harmonised standards to start.</strong> Design against IEC 62443 and document the reasoning, and the standard, when it lands, will be a mapping exercise rather than a redesign.</li>
</ol>
<h2 id="what-we-are-not-claiming">What we are not claiming</h2>
<p>Two limits, stated because compliance writing tends to blur them.</p>
<p>We are describing published dates, not predicting enforcement, because how aggressively national market surveillance authorities act in the first year of the Machinery Regulation is genuinely unknown, and anyone telling you otherwise is guessing, so the dates are certain while the intensity is not.</p>
<p>None of this is legal advice, and the two regulations interact in ways the Commission itself has undertaken to clarify through guidance, so if your product sits near a category boundary (a safety component, a machine embedding machine learning, a product that is arguably both machinery and a product with digital elements) that boundary is worth a specialist&#39;s opinion rather than a blog article&#39;s.</p>
<p>What we can say with confidence is narrower and, we think, more useful, because from 2027 the cybersecurity of a machine&#39;s safety functions is a CE-marking matter, opening a machine to secure it carries a legal consequence it did not carry before, and the installed base that will never comply on its own terms is the part of the estate that needs governing from outside. That, in a few lines, is what the 2027 cybersecurity legislation asks of you.</p>
<p><em>Last reviewed: August 2026. Regulatory timetables and standardisation status change, so verify against <a href="https://eur-lex.europa.eu/">EUR-Lex</a> and your notified body before relying on any date here.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>How to Navigate NIS2 Compliance for Your Business</title>
      <link>https://trust-buster.com/blog/how-to-navigate-nis2-compliance-for-your-business</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/how-to-navigate-nis2-compliance-for-your-business</guid>
      <pubDate>Wed, 19 Aug 2026 11:54:07 GMT</pubDate>
      
      <description>Start with the inventory, then keep walking past the server room, because NIS2 compliance begins where most programmes stop, at the door of the workshop.</description>
      <content:encoded><![CDATA[<p class="lead"><strong>Start with the inventory, then keep walking past the server room</strong></p>
<p><em>Achieving NIS2 compliance, which means meeting the obligations of the EU directive on the security of network and information systems, is essential for businesses operating within the EU, particularly for those managing critical infrastructure, and the first step is a thorough inventory of your IT <strong>and OT</strong> assets, to understand where you are exposed.</em></p>
<p><em>That first step is where most programmes quietly go wrong, because the inventory, even when it is done well, stops at the door of the workshop.</em></p>
<h2 id="the-assets-that-were-never-on-anyone-s-list">The assets that were never on anyone&#39;s list</h2>
<p>Run the inventory properly and you will produce two lists.</p>
<p>The first is familiar, because it holds servers, applications, endpoints, identities and network equipment, and every item has an owner, a patch cycle and, usually, an access policy.</p>
<p>The second list is the one nobody has ever compiled. Machines, electrical cabinets, test benches, furnaces, presses, autoclaves, drives, in other words equipment where unauthorised use is not a confidentiality problem but a safety, counterfeiting or sabotage problem, and for most of these you cannot produce a list of who is entitled to operate them, and you cannot produce a log of who actually did.</p>
<p>This list is not missing through negligence, because it fell outside everyone&#39;s scope, too physical for IT and too digital for maintenance, and in a manufacturing, energy or water entity it is frequently longer than the governed-application list it sits beside.</p>
<p>Compiling it is, in our view, the most useful thing you can do in your first quarter of NIS2 compliance work, because everything downstream (risk analysis, access control policy, incident classification, supplier clauses) is scoped by it.</p>
<h2 id="what-article-21-2-actually-asks-of-you">What Article 21(2) actually asks of you</h2>
<p>NIS2 lists ten categories of risk-management measure, and three of them reach the shop floor directly, so they are worth reading with an OT asset in mind rather than a server.</p>
<ul>
<li><strong>(c) business continuity and crisis management.</strong> An unauthorised start-up mid-cycle is a continuity event, not just a security one.</li>
<li><strong>(i) human resources security, access control policies and asset management.</strong> Note the pairing, because access control policy and asset management are named in the same breath, which is precisely the pairing that breaks when the asset is a press with a key switch.</li>
<li><strong>(j) multi-factor authentication or continuous authentication solutions.</strong> Written for logical access, but nothing in the text confines it there, and continuous authentication is exactly what an energised machine needs and almost never has.</li>
</ul>
<p>The directive also frames the whole of Article 21(2) as an all-hazards approach, aimed at protecting network and information systems <em>and the physical environment of those systems</em>, and that phrase is doing more work than most NIS2 compliance programmes give it credit for.</p>
<h3 id="the-word-physical-is-nearly-absent-from-article-21-which-is-">The word &quot;physical&quot; is nearly absent from Article 21, which is why it gets missed</h3>
<p>Read Article 21 looking for physical security and you will barely find it, and that absence has led a great many programmes to conclude, reasonably enough, that NIS2 is about networks and leaves doors and machines to somebody else.</p>
<p>The Commission does not read it that way, because its Implementing Regulation (EU) 2024/2690, adopted in October 2024, translates Article 21(2) into more than a hundred auditable technical requirements, and one of its thirteen annex sections is devoted to physical and environmental security, covering supporting utilities, perimeter and physical access controls, monitoring, and periodic review of those measures. The regulation attaches this to points (c), (e) and (i) of Article 21, even though physical security is not named there.</p>
<p>Be clear about scope, because this is where commentary gets sloppy, since that implementing regulation binds a specific subset of entities, namely DNS and TLD operators, cloud and data centre providers, CDNs, managed service and managed security providers, online marketplaces and search engines, social platforms and trust service providers, so if you run a factory it does not apply to you.</p>
<p>What it does give you is the Commission&#39;s own reading of what Article 21(2) means in practice, written down and directly applicable somewhere in the Union, and ENISA&#39;s <a href="https://www.enisa.europa.eu/publications/nis2-technical-implementation-guidance">technical implementation guidance</a>, published in June 2025, mirrors its structure and includes the kind of evidence an auditor will expect to see. When your national supervisor or your customer&#39;s auditor asks what &quot;appropriate measures&quot; looks like for access control and asset management, this is, as far as we know, the most authoritative answer currently in writing, and it treats the power supply and the physical perimeter as in scope.</p>
<p>Planning as though physical access to machines is outside NIS2 is a bet against the direction of travel.</p>
<h2 id="incident-response-planning-tested-against-the-shop-floor">Incident response planning, tested against the shop floor</h2>
<p>A robust incident response plan is a cornerstone of NIS2 compliance, because your plan should set out clear procedures for detecting, reporting and mitigating incidents, and the directive puts a clock on it.</p>
<h3 id="the-24-72-one-month-cascade">The 24 / 72 / one-month cascade</h3>
<p>Article 23 requires an early warning within 24 hours of becoming aware of a significant incident, a fuller notification within 72 hours including an initial assessment of severity and impact, and a final report within one month, and regular exercises are what make those windows survivable, because a plan that has never been rehearsed is a document, not a capability.</p>
<h3 id="you-cannot-report-what-you-cannot-attribute">You cannot report what you cannot attribute</h3>
<p>Here is where OT assets break the plan, and it is worth walking through before it happens rather than during.</p>
<p>At 22:40 on a Friday, a five-axis machine runs for forty minutes and produces eleven parts, and on Monday someone notices, so now start the clock. The chain that should be able to answer what follows is the one we describe in <a href="/blog/iam-at-the-end-of-the-power-cord-part-1">IAM at the end of the power cord</a>.</p>
<p>Within 24 hours you owe an early warning, and within 72 hours you owe an initial assessment of severity and cross-border impact, but to produce either you need to answer three questions, who started it, under what entitlement, and was this unauthorised use or a scheduling error.</p>
<p>If the only record is the machine&#39;s hour counter (unattributed, unsigned, reset at the last service) you cannot answer, so you will file a notification that says something happened and you are investigating, and you will still be saying that a month later in the final report. That is not a reporting failure caused by bad process, it is a reporting failure caused by a missing log that was never designed to exist.</p>
<h3 id="what-a-usable-ot-incident-record-looks-like">What a usable OT incident record looks like</h3>
<p>The requirements are unlike your SIEM&#39;s, the tool that collects and analyses your IT logs, because volume is low, a few dozen events per machine per month, while unit value is very high, and the consumer is rarely a real-time detection engine, since it is a supervisor, an insurer, a prime contractor or a regulator, reading it much later.</p>
<p>That points to four properties:</p>
<ul>
<li><strong>Local.</strong> The network was very likely down, or absent by design, at the moment that matters.</li>
<li><strong>Timestamped to a trusted clock.</strong> An OT clock that is twelve hours out or still on winter time makes every other property worthless.</li>
<li><strong>Chained and signed</strong>, so that a missing entry is visible rather than merely absent.</li>
<li><strong>Exportable</strong> without depending on any central infrastructure.</li>
</ul>
<p>Attribution is the hard part, and it is an identity problem before it is a logging problem, because a log can only record the identity that the enforcement point actually established, which is why access control and evidence have to be designed together rather than in sequence. We have written separately about what a proof log proves and what it does not.</p>
<h2 id="supply-chain-the-pressure-that-arrives-before-the-regulator-">Supply chain: the pressure that arrives before the regulator does</h2>
<p>Article 21(2)(d) makes supply chain security an explicit obligation, which means your customers&#39; NIS2 compliance obligations become your questionnaire.</p>
<p>In practice this is the pressure that reaches a plant manager first, because automotive, aerospace, energy and defence primes are already sending cyber clauses down to tier-2 and tier-3 suppliers, and those clauses increasingly ask for things the shop floor cannot produce, such as a list of who may operate a given asset, evidence that qualifications were current, or a record of physical interventions by third-party technicians.</p>
<p>Answering &quot;we don&#39;t track that&quot; is a commercial problem long before it is a regulatory one.</p>
<h2 id="three-places-the-standard-playbook-does-not-transfer">Three places the standard playbook does not transfer</h2>
<p>Briefly, because each deserves its own treatment.</p>
<p><strong>Fail-closed is not automatic.</strong> In IT, denying on doubt (what is usually called failing closed) is the default, but cutting power to a ventilation system or a cooling unit is a safety decision before it is a security one, so the meaning of &quot;deny&quot; has to be set equipment by equipment, jointly with operations and HSE.</p>
<p><strong>Connectivity is not an assumption.</strong> A decision point that calls a remote service is a decision point that abstains during an outage, precisely when risk peaks, so decisions have to be rendered locally and remain nominal, not degraded.</p>
<p><strong>Most legitimate subjects are not yours.</strong> Temps, tier-2 subcontractors, OEM engineers flying in from another continent, and you have to verify an entitlement and tie it to a work order without ever having provisioned the person, often with no directory on the supplier&#39;s side.</p>
<h2 id="where-transposition-actually-stands-and-why-it-does-not-chan">Where transposition actually stands, and why it does not change your deadline</h2>
<p>The directive required national law by 17 October 2024, and four member states met it. In November 2024 the Commission opened infringement proceedings against 23 of them, and issued reasoned opinions to 19 in May 2025. By mid-2026 the large majority had adopted transposing legislation, but a handful, including Spain and the Netherlands, still had not completed the process, and in July 2026 the Commission referred France, Ireland, Spain and the Netherlands to the Court of Justice, asking for financial penalties.</p>
<p>It is tempting to read that as breathing room, but it is not, for three reasons, because your customers&#39; clauses do not wait for a parliament, because entities operating across several member states are already in scope somewhere, and because the obligations arriving in late-transposing states are, by design, the same ones. The organisations that used the delay to compile the OT inventory will be ready, while the ones that used it to wait will be doing the same work under a supervisor&#39;s timetable rather than their own. The rest of the 2027 calendar, including the Machinery Regulation and the Cyber Resilience Act, is covered in <a href="/blog/emerging-trends-in-cybersecurity-legislation-what-to-expect-in-2027">what to expect in 2027</a>.</p>
<h2 id="a-practical-sequence-for-nis2-compliance-on-the-shop-floor">A practical sequence for NIS2 compliance on the shop floor</h2>
<ol>
<li><strong>Extend the inventory to OT assets with no access policy.</strong> Not every machine, but those where unauthorised use carries physical, safety or regulatory consequence.</li>
<li><strong>Define the meaning of &quot;deny&quot; for each of them,</strong> with operations and HSE in the room, because this is the hardest deliverable and it conditions everything else.</li>
<li><strong>Require an enforcement point that can decide offline.</strong> In our experience this single criterion eliminates unsuitable options faster than any feature comparison.</li>
<li><strong>Design the evidence trail with the access control,</strong> not after it, because you can only log the identity you actually established.</li>
<li><strong>Rehearse the 24-hour clock against an OT scenario,</strong> not an IT one, since that is where the plan will break, and better in an exercise.</li>
</ol>
<p>And one thing not to do, <strong>do not connect equipment to the network for the sole purpose of governing it</strong>, because creating an attack surface in order to monitor the attack surface you just created is a poor trade. Physical access control does not require the machine to be smart, it requires the power outlet to be.</p>
<h2 id="what-this-does-not-give-you">What this does not give you</h2>
<p>Compliance work invites overstatement, so to be plain about the limits, controlling and logging access at the power supply answers the access control, asset management and evidence parts of Article 21(2), but it does not answer network segmentation, vulnerability handling, backup and recovery, cryptography policy or supplier due diligence. It is one column of the NIS2 compliance matrix, not the matrix.</p>
<p>It happens to be the column that is empty on almost every site, which is why it is worth starting there, and why it is worth being honest about where it stops.</p>
<p><em>Last reviewed: August 2026. NIS2 transposition and enforcement statuses change, so verify against your national competent authority before relying on any date here.</em></p>
]]></content:encoded>
    </item>
    <item>
      <title>Inside Trust-Buster: Building a Platform that Meets Legal and Security Challenges</title>
      <link>https://trust-buster.com/blog/inside-trust-buster-building-a-platform-that-meets-legal-and-security</link>
      <guid isPermaLink="true">https://trust-buster.com/blog/inside-trust-buster-building-a-platform-that-meets-legal-and-security</guid>
      <pubDate>Wed, 19 Aug 2026 11:54:07 GMT</pubDate>
      
      <description>How we actually decide inside Trust-Buster, including the legal and security decision we had to reverse when two European components changed hands.</description>
      <content:encoded><![CDATA[<p class="lead"><strong>How we actually decide, including the decision we had to reverse</strong></p>
<p><em>The Ops-Devices platform was designed to address technical needs and the legal and security requirements arriving with NIS2, the CRA and the Machinery Regulation, and that much is on the front page of every vendor in this space, so it tells you very little.</em></p>
<p><em>What follows is the part that is harder to write, because it covers how the legal and security decisions get made, what we refuse to claim, and the architectural commitment we had to unwind when the ground moved under it.</em></p>
<h2 id="we-keep-a-register-of-the-alternatives-we-rejected">We keep a register of the alternatives we rejected</h2>
<p>Most engineering teams document what they built, whereas we also document what we did not build, and why, a practice that is now common and for which resources such as the <a href="https://github.com/architecture-decision-record/architecture-decision-record">Architecture Decision Records</a> are worth reading and using.</p>
<p>Every significant architectural choice goes into a register that records four things, the option retained, the options discarded, the reasoning at the time, and the part that matters most to us, the re-evaluation trigger. That last field is a written statement of what would have to change in the world for the decision to be reopened.</p>
<p>This started as ordinary R&amp;D discipline, and it turned out to have two other uses we did not anticipate.</p>
<p>The first is regulatory, because both the CRA and the Machinery Regulation ask manufacturers to demonstrate that security was considered by design, not asserted afterwards, and a dated register of rejected alternatives with reasoning attached is a far better answer to that legal and security question than a policy document written the month before the audit.</p>
<p>The second is honesty enforcement, because it is uncomfortable to write down a re-evaluation trigger, since you are committing in advance to a condition under which you will have to admit that a decision no longer holds, which brings us to the one that fired.</p>
<h2 id="the-decision-we-had-to-reverse-on-our-hardware-trust-chain">The decision we had to reverse on our hardware trust chain</h2>
<p>Our early hardware architecture was anchored on a European trust chain, because the cellular platform we had selected combined a microcontroller with a modem carrying a secure enclave whose IP, meaning the licensed design of the protected area of the chip where the keys live, came from Tiempo Secure, a Grenoble-area company founded in 2007, whose TESIC RISC-V secure element had passed Common Criteria EAL5+ evaluation with the French ITSEF laboratory SERMA, and which had been selected for a French State-backed sovereign post-quantum security microprocessor programme. The modem line came from Sequans, a French supplier of cellular semiconductors, so European board, European modem, European enclave IP, a chain we could describe end to end and defend in front of a defence procurement officer.</p>
<p>Then both pieces changed hands.</p>
<p>In August 2024 Qualcomm agreed to acquire Sequans&#39; 4G IoT technology assets, including employees and licences and the Monarch product line, and the transaction completed the following month for around $200 million, with Sequans retaining a perpetual licence to continue using and commercialising the technology. Tiempo Secure was subsequently acquired by Qualcomm as well, a transaction confirmed publicly by its co-founder rather than announced by press release, and now reflected on Tiempo&#39;s own website.</p>
<p><strong>What this did and did not change.</strong> It is worth being precise, because the loose version of this story is more dramatic than the true one. The technology did not stop working, Sequans continues to operate and holds a perpetual licence, Tiempo&#39;s engineering team remains in Montbonnot, and nobody&#39;s chip turned into a backdoor overnight.</p>
<p>What changed is narrower and, for us, decisive, because the sentence we had been writing was no longer true. We had described a fully European hardware trust chain, and after the acquisitions the enclave IP and a substantial part of the modem technology sat under US corporate ownership. For an industrial customer, that is a footnote, but for a defence customer evaluating supply-chain provenance and the digital sovereignty of its suppliers, it is the whole question.</p>
<p><strong>What we did.</strong> We deprecated that platform as our sovereignty anchor and moved the anchor elsewhere, to an STM32 microcontroller, a Nordic cellular bearer, and a discrete secure element sourced separately rather than inherited through a modem. The prototyping path now runs on a Nordic development platform with an STSAFE secure element, with a higher-assurance discrete element on the production path.</p>
<p><strong>What we learned</strong>, and this is the part worth generalising, fits in one sentence, <em>sovereignty is not a property of a component, it is a property of a claim you have to keep true.</em> Ownership of semiconductor IP changes hands on a timescale much shorter than the service life of an industrial machine, so any architecture that stakes its differentiation on the nationality of a supplier needs an explicit re-evaluation trigger attached to it, because that trigger will eventually fire, and the alternative to noticing it yourself is a customer noticing it for you.</p>
<p>Our register now carries that trigger on every sourcing decision that carries a provenance claim.</p>
<h2 id="three-legal-and-security-claims-we-refuse-to-make">Three legal and security claims we refuse to make</h2>
<p>Compliance-adjacent marketing has a gravitational pull toward overstatement, and these are the three places we have decided in advance not to follow it, written down so that a reader can hold us to them.</p>
<p><strong>EAL5+ is inherited, not obtained.</strong> When we reference a Common Criteria EAL5+ evaluation, it belongs to a secure element component and to the laboratory that evaluated it, because Ops-Devices holds no such certification and does not claim one. A product containing a certified component is not a certified product, and the distinction matters precisely to the customers who care about certification at all.</p>
<p><strong>Radio signal strength is a gate, not a proof.</strong> Our system uses received signal strength as an entry condition, so below a threshold the exchange does not begin, but it is not an authentication factor, because signal strength can be amplified, relayed and spoofed, and no amount of filtering turns it into a distance measurement. Time-of-flight ranging over ultra-wideband does that properly, but we do not have it today, and we say so rather than implying that careful thresholding is equivalent.</p>
<p><strong>We cannot claim exhaustive detection.</strong> No radio-frequency monitoring system can demonstrate the absence of a signal it did not see, so detection is a corroborating input to a decision, never the anchor of one, and any vendor claiming complete detection coverage in an uncontrolled radio environment is describing an ambition, not a measurement.</p>
<p>Each of these makes a slide slightly less impressive, but each of them also survives contact with a customer&#39;s technical evaluation, which the impressive version does not.</p>
<h2 id="collaboration-and-innovation">Collaboration and innovation</h2>
<p>Our development process does involve continuous engagement with the outside world, but it is worth describing what that actually consists of, because the generic version of this sentence claims more than we can support.</p>
<p><strong>We design against published standards, and we do not sit on the bodies that write them.</strong> IEC 62443 shapes our architecture, the CRA and the Machinery Regulation shape our documentation, and IEEE 802.1AR, ISO 15118 and the IETF&#39;s constrained-environment work inform specific design choices. We are readers of those standards and, where relevant, we prepare structured responses for procurement teams who cite them, but we do not hold seats on standards committees or regulatory bodies, and describing our work as collaboration with regulators would be an overstatement of a different kind of relationship.</p>
<p><strong>We build on open source, deliberately and by default.</strong> Firmware on open frameworks, published schematics in an open EDA format, off-the-shelf modules with documented protocols rather than bespoke hardware wherever a standard part will do, and this is not ideology, because it is the only way a small team keeps a supply chain it can actually audit, and it is what makes the provenance question above answerable at all.</p>
<p><strong>The innovation is confined to one layer.</strong> Assembling proven components and open frameworks is not innovation, and we do not present it as such. What we keep in-house is the chain of trust that ties those components together, meaning local policy decision without connectivity, ephemeral signed capabilities, an evidence trail designed for retrospective proof rather than real-time detection, and enforcement that sits <a href="/blog/ot-retrofit-addons-instead-of-upgrades">at the boundary of a machine rather than inside it</a>. That is a narrow layer, and as far as we know it is also the layer nobody else in the competitive landscape occupies, which is why it is the only part worth building ourselves.</p>
<h2 id="what-the-platform-answers-and-what-it-does-not">What the platform answers, and what it does not</h2>
<p>A platform that claims to deliver compliance is describing something that does not exist, because compliance is an organisational outcome, and a product can only close specific legal and security requirements within it.</p>
<p>The Ops-Devices platform addresses a defined column of that matrix, access control and asset management for equipment that has no digital path, evidence of physical intervention that survives an offline site and an eighteen-month gap, and enforcement that does not require modifying the machine.</p>
<p>It does not address network segmentation, vulnerability handling, backup and recovery, cryptography policy across your estate, supplier due diligence, staff training, or incident response process. Those are real obligations under <a href="/blog/how-to-navigate-nis2-compliance-for-your-business">NIS2 Article 21(2)</a>, and anyone telling you a single product covers them is selling you a compliance narrative rather than a control.</p>
<p>The reason we start where we start is not that our column is the most important one, it is that on almost every site we visit, it is the one that is completely empty, with no list of who may operate the asset, no record of who did, and no mechanism that could produce either.</p>
<h2 id="why-we-write-any-of-this-down">Why we write any of this down</h2>
<p>There is a straightforward commercial argument against publishing a reversed architectural decision, an inherited certification caveat and a list of things a product does not do, because it makes the story less clean.</p>
<p>We do it because of who evaluates this category, since the people who buy access control for industrial and defence sites are, by profession, people who look for the gap between the claim and the artefact, and a vendor who names their own limits first is not being modest, they are removing the most productive line of questioning from the evaluation and leaving only the substance.</p>
<p>And we do it because of the register, because once you have written down the condition under which a decision stops being true, you have to do something when it happens. That is an uncomfortable discipline, and it is the closest thing we have to a guarantee that what we tell you next year on legal and security matters will still be defensible.</p>
]]></content:encoded>
    </item>
  </channel>
</rss>
