The core downside: Context vs. guidelines
Conventional SAST instruments, as we all know, are rule-bound; they examine code, bytecode, or binaries for patterns that match identified safety flaws. Whereas efficient, they typically fail in relation to contextual understanding, lacking vulnerabilities in advanced logical flaws, multi-file dependencies, or hard-to-track code paths. This hole is why their precision charges and the proportion of true vulnerabilities amongst all reported findings stay low. In our empirical research, the extensively used SAST instrument, Semgrep, reported a precision of simply 35.7%.
Our LLM-SAST mashup is designed to bridge this hole. LLMs, pre-trained on large code datasets, possess sample recognition capabilities for code habits and a information of dependencies that deterministic guidelines lack. This enables them to motive in regards to the code’s habits within the context of the encircling code, related recordsdata, and the complete code base.
A two-stage pipeline for clever triage
Our framework operates as a two-stage pipeline, leveraging a SAST core (in our case, Semgrep) to determine potential dangers after which feeding that info into an LLM-powered layer for clever evaluation and validation.