Researchers have long considered the Stuxnet attacks on Iran’s nuclear centrifuges in Natanz to be the opening chapter of state-sponsored cyber sabotage.
As it turns out, at least five years before Stuxnet became public in 2010, somebody had developed an equally potent cyber weapon, one capable of injecting near-imperceptible errors into high-precision mathematical computations to gradually undermine and sabotage systems and applications that rely on their results.
Researchers at SentinelOne who discovered the previously undocumented malware framework, which they are tracking as fast16, say it represents the earliest example yet of a cyber tool designed explicitly for sabotaging “ultra expensive high-precision computing workloads of national importance like advanced physics, cryptographic, and nuclear research workloads.”
“The discovery of fast16 rewrites our understanding of what a cyber weapon can do, as well as when nation-state cyber sabotage operations matured to the level of becoming a serious threat to critical infrastructure,” says SentinelOne researcher Vitaly Kamluk in comments to Dark Reading.
Rewriting Notions of a Cyberweapon
Fast16’s function was to quietly corrupt mathematical outputs of engineering and scientific software by introducing tiny systematic errors that would be nearly impossible to detect without running independent calculations on a completely separate, uninfected system.
SentinelOne likened fast16’s delivery mechanism to a “cluster munition” that could drop multiple “wormlets” that would then distribute the main payload to as many machines as possible in a target environment by looking for and exploiting vulnerabilities in them.
Fast16 marks a major turning point in the history of cyber weapons, Kamluk says. “Despite its twenty-year vintage, we have yet to discover another malware specifically designed to compromise high-precision mathematical calculations in this way.”
A Fortunate Find
SentinelOne researchers uncovered fast16 while attempting to trace the earliest meaningful use of an embedded Lua VM in Windows malware. Lua is a scripting language that organizations use to extend an application’s functionality. SentinelOne had observed how the authors of highly sophisticated malware frameworks such as Flame, Flame 2.0, PlexingEagle, and Project Sauron consistently embedded a Lua scripting engine to add modularity to their tools and wanted to see how far back the practice went.
What they discovered was fast16, with components dating back to 2005, well before the earliest known use of Stuxnet, widely regarded as the first known deployment of a cyber weapon in a geopolitical context. Their analysis of fast16 showed it to be the first-ever Lua-based network worm targeting high-precision calculation software.
The name “fast16” appears in a document the ShadowBrokers group leaked in 2016 regarding the National Security Agency’s offensive cyber weapons. But SentineOne did not attribute the malware to NSA or any other entity.
Remarkably, someone had uploaded the malware to VirusTotal more than a decade ago, where it has remained almost undetected. Only one engine on VirusTotal classifies the tool as generally malicious, but even that is with moderate confidence, SentinelOne said. While that VirusTotal result may appear concerning, Kamluk noted that fast16 “is genuinely an old piece of malware” that only runs in an “environment that is largely obsolete.”
“Frankly, we believe we were fortunate simply to pick up the trace, as it was surrounded by misleading false vectors that could easily have led other researchers to an incorrect hypothesis without proper validation.”
Targeted Software Suites
The researchers identified three software suites that fast16 likely targeted: LS-DYNA 970, PKPM, and the MOHID hydrodynamic modeling platform, all used for scenarios like crash testing, structural analysis, and environmental modeling. SentinelOne identified LS-DYNA as software that Iran is reported to have used in computer modeling relevant to its nuclear weapons development program, suggesting it might have been a target even before Stuxnet.
However, researchers are unsure if the authors — most likely state actors — ever deployed the weapon, what its intended targets are, or impact it would have in an actual attack scenario.
“As for geopolitical contexts and nation states, the malware has no specific reference about where it was meant to be deployed,” Kamluk says. “The targeted software could pop up anywhere.”
Still, Kamluk assesses fast16 as likely the work of a nation-state actor. “Patching software that performs high-precision physical process simulation is beyond the scope of a typical developer,” he says. “It requires intimate familiarity with the specific subject field to create subtle yet meaningful sabotage alterations.”
An Attack Vector Remains Relevant
Considering the software’s age, it is extremely difficult to know if any organizations fell victim to fast16. So, it is only possible to speculate on the possible outcomes of a fast16 attack.
Considering it was written for a different generation of systems, fast16 is incapable of running on modern systems, Kamluk says. The malware runs only on uniprocessor Windows XP systems, an environment that is now largely obsolete. Even in rare instances when such legacy systems persist in old laboratories, installing modern security software on them is often impossible, he notes.
“[But] the underlying attack vector remains highly relevant. High-precision calculations, whether used in financial trading, AI model training, or various simulation software, could still be the target of a similar, but modernized threat today.”
SentineOne has published Yara rules that organizations can use to check older systems or data archives.
“The true significance of the fast16 discovery lies in identifying this novel and unusual cyber sabotage attack vector itself,” Kamluk says.
