Vulnerability Pattern

Vulnerability pattern research focuses on identifying and characterizing recurring structures in vulnerable code to improve automated software vulnerability detection. Current efforts utilize machine learning, particularly deep learning models like recurrent neural networks and graph attention networks, along with techniques like multi-task learning and contrastive learning, to analyze code features (e.g., program slices, code property graphs) and predict vulnerabilities at the function or even statement level. This work is significant for enhancing software security by enabling more accurate and efficient vulnerability detection, potentially reducing the risk of exploitation and improving software reliability.