Introducing AutoPatchBench: A Benchmark for AI-Powered Security Fixes

What this post added

Introduces AutoPatchBench, a benchmark for evaluating AI program repair systems specifically for fuzzing-identified vulnerabilities. It includes 136 C/C++ vulnerabilities with verified fixes from the ARVO dataset, along with a rigorous automated verification process involving fuzzing and white-box differential testing. This aims to standardize the evaluation of AI-driven security fixes and accelerate progress in automated vulnerability repair.