900 Vulnerability Hypotheses and 12 Billion Tokens

The author shares their experience researching Rust security after starting a project to develop a high-performance network application. Faced with the need to audit code for vulnerabilities, they built a custom research pipeline called "rust-in-peace." The process involved analyzing errors made by both human developers and modern Large Language Models (LLMs). The research included creating signatures for SAST tools, finding bugs in popular libraries, and even contributing a patch to the Linux kernel. The author details how using agents and automated verification systems helped identify hundreds of hypothetical vulnerabilities using 12 billion tokens for analysis. The article provides a deep technical dive into how modern automation tools can be applied to ensure memory safety and find critical bugs in Rust code, while also summarizing the author's presentation at the ZeroNights conference.