oss-fuzz-gen

The OSS-Fuzz-Gen framework leverages large language models (LLMs) to automatically generate fuzz targets for C, C++, Java, and Python projects, enhancing the bug detection capabilities within the OSS-Fuzz platform. It evaluates these targets based on compilability, runtime crashes, and coverage metrics, showcasing significant improvements in code coverage for various open-source projects. Notably, it has successfully identified new vulnerabilities, demonstrating the efficacy of AI in enhancing software security testing.