The classifier judged this item not relevant to AI security, so it is left out of lists, feeds, the API and dataset releases. If that is wrong, report a correction on this page.
InfoResearchPreprint
VulnSAGE: Adjudicating SAST Reports by Connecting LLM Reasoning to Program Structure
- Published
- Record updated
Summary
VulnSAGE is a system that adjudicates SAST reports by linking LLM reasoning to program structure. On 187 labeled reports it rejects 89 of 97 false alarms and retains 85 of 90 real defects. Applied to 45,440 alerts on the 200 most-starred C/C++ GitHub projects, it reproduces 104 defects at runtime, and maintainers have confirmed 24 of the 26 they responded to.