Comparison of Static Application Security Testing Tools and Large Language Models for Repo-Level Vulnerability Detection
Zhou, X., Tran, D.-M., Le-Cong, T., Zhang, T., Irsan, I. C., Sumarlin, J., Le, B., & Lo, D. (2024). Comparison of static application security testing tools and large language models for repo-level vulnerability detection. arXiv preprint arXiv:2407.16235.
This study compares 15 SAST tools with 12 open-source LLMs across Java, C, and Python repositories. Findings indicate that while SAST tools have lower false positives, they detect fewer vulnerabilities compared to LLMs. Combining both approaches can enhance detection accuracy.