Skip to content

The Novel Was Right

· 9 min read · Updated

Lire en français
The Novel Was Right

A piece of technical speculative fiction usually ages badly, because its details end up looking quaint, but Blackout 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.

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.

The book

Blackout: Tomorrow Will Be Too Late 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 Zero two years later.

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.

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'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.

What was already true when he was writing, from Stuxnet to Shodan

Elsberg invented very little, and he does not pretend otherwise, because two things were already public when he wrote.

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's idea.

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'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 academic study at Cambridge 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.

What reality has added since

This is where the fiction gets overtaken.

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 report 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.

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's and not on a defence, as the FBI notification 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.

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'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.

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.

And across Europe more broadly, in January 2026 the French national cybersecurity agency published a recommendation 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.

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's manual.

What the novel exaggerates, and why saying so matters

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.

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.

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.

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.

The novel was right, but not where one expects

Closing the book, the temptation is to conclude that large infrastructure needs better protection, and that is true, and largely under way.

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'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.

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, someone standing physically in front of the machine, 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.

So, is it worth reading?

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.