KNIGHTER

Summary from the TLDR InfoSec newsletter

KNighter is an automated static analysis tool from UIUC that uses Large Language Models to synthesize custom code checkers by learning from historical patch commits in C/C++ codebases like the Linux kernel. The tool employs a three-stage pipeline of generation, refinement, and triage to create LLVM-based checkers that identify bug patterns, focusing on finding real-world vulnerabilities in large-scale systems. It integrates with OpenAI, Claude, Google, and DeepSeek models to automatically generate static analysis rules from git commit history, enabling security teams to discover previously unknown bugs based on patterns from past fixes.