No, it is not a cyberpunk science-fiction story: this new thing just appeared, for real!
An AI-driven worm that propagates across a heterogeneous network by parasitically acquiring computational resources for autonomous reasoning.
a fundamentally new cybersecurity threat: an adaptive computer worm that devises target-specific attack strategies to gain control of machines and spread across networks. Each compromised machine becomes part of the worm’s own infrastructure, providing compute or reach for further attacks
The research:
Traditional worms can be stopped by patching the specific vulnerability they exploit. Our adaptive worm cannot be stopped this way: it uses a recursive reasoning loop to detect and exploit diverse vulnerabilities as it propagates.
What’s more, because this design is built using a small model that runs on a single machine, the economics of cyberattacks are about to radically shift
EDIT: removed one comment that was considered offensive
There are LLM systems that can find previously unknown vulnerabilities, but they are big models requiring significant hardware and of course access to all the source.
Without that sort of thing a LLM based worm can only go through previously catalogued exploits. Which is something one could do by just bundling a heap fo exploits into a single program with a bunch of switch statements to determine which ones to try in different situations.
This “AI Worm” will do less damage than a typical worm that uses a single vulnerability in a recent version of Windows.
Another angle though would be … an AI worm could more readily adjust to attempts to counter it. It would be more agile and more responsive to the real-time actions of either automated security software or security staff or both. In the extreme both attacker and defender could be AI. However, at the end of the day, it all starts with a vulnerability.
I have sleepless nights about it but.. no. Its nothing new.
If you have a bad time like humans could exploit your systems with a huge pile of time, energy and money. Computers could do it too right now with a shift of time and data.
Try to have a chain of computers and ssh connections and later (quantum encrypted connections) between your systems or do and allow connections only on a very small random time frame. And your systems will be safe if you can keep the offline documents secret and share it only with friends.
But yes for so many systems out there it will be a mess.
Oh and have only air disconnected offline systems too.
