Vulnerability Management is being reshaped by the rise of LLMs and generative AI agents, challenging the long-standing dominance of human-led SAST. While SAST offers scalable, rule-based detection with comparatively lower false positives, it often struggles with semantic and cross-file context; conversely, LLMs can generalize across vulnerability descriptions and code patterns but suffer from hallucinations and high false-positive rates when used alone. Currently, surveys show evidence from recent repo-level comparisons and highlight the converging trend: hybrid neuro-symbolic approaches that use LLMs for hypothesis generation and static/taint/formal methods for validation, improving reliability over conventional pipelines.
Short Bio:
Cătălin is a first-year PhD Student at West University of Timisoara with interests in Cybersecurity. His doctoral research topic is Automated Vulnerability Discovery powered by Large Language Models. Catalin has a Bachelor’s degree in Computer Science and a Master’s Degree in Cybersecurity, both from West University of Timisoara. He is currently employed as a Senior Security Engineer at Manage Now, focusing on Endpoint Protection, Data Loss Protection, Email Security, Secure Web Gateways, and Account Password Security.
