Enhancing Java Fuzzing with LLMs: Challenges and Solutions in OSS-Fuzz

The blog post covers the development of Java fuzz harness synthesis using large language models (LLMs) in the OSS-Fuzz project, which aims to automate fuzzing for open-source software. Expanding from C/C++ to Java, the authors face unique challenges, such as managing object lifecycles, handling exceptions, ensuring proper resource management, and selecting suitable target methods. To address these issues, they provide LLMs with detailed prompts that guide the generation of fuzzing harnesses, ensuring effective object instantiation, appropriate exception handling, and resource closure. The results show that of the 592 targets identified, 280 harnesses built successfully, with 102 achieving code exploration.


 https://blog.oss-fuzz.com/posts/introducing-java-auto-harnessing

Comments

Popular posts from this blog

Endor Labs Announces Integrated SAST Offerings

The Hidden Cost of DevSecOps: Time and Financial Burden of Security on Developers

OWASP Releases Enhanced Dependency-Check Tool with Advanced Tagging and Policy Management Features