Combining LLM and Dynamic Analysis for Universal Testing of Static Application Security Testing Tools
LLM4SAST finds 66 SAST-tool faults, with 32 confirmed or fixed by developers.
LLM4SAST uses large language models and dynamic feedback to find faults in static application security testing tools across programming languages. Metamorphic relations guide variant generation, while dynamic-feedback selection and program-differencing clustering reduce bad variants and duplicate reports. It identified 66 unique faults across six SAST tools, 32 of which were confirmed or fixed, eliminated 47.8% of false positives, and found 3.4 times more faults using only 39.8% of the variants required by earlier methods.
- LLM4SAST generates multi-language variants using metamorphic relations.
- Heuristics cut false positives by 47.8 percent.
- Found 66 unique faults in six SAST tools; 32 confirmed or fixed.
- Detects 3.4 times more faults with 39.8 percent of prior variants.
From the original article
1 min read at arxiv.orgStatic application security testing (SAST) tools help detect potential software vulnerabilities, but they themselves contain faults undermining reliability and usability. Automated testing can address these issues, yet current techniques suffer from limited program transformation diversity and lack mechanisms to verify generated variants, leading to reduced fault detection and high manual effort.
By Huaien Zhang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.