Autonomous Exploitation at Scale
With the advent of powerful, open-source, and lightweight large language models, the cost barriers that typically prevent targeted attacks – with people manually scanning, infecting, and exploiting – are vastly reduced. This talk presents a self-propagating worm which autonomously detects devices, enumerates their vulnerabilities and services, checks for exploits which could work on them, and exploits them and spreads to them, forming a distributed network. It discusses the limitations, design decisions, and tradeoffs made in the development.
