Code Representation

We propose a graph neural network vulnerability mining system based on hybrid semantics, which constructs a composite semantic code property graph for code representation based on the causes of vulnerabilities.

This project studies the security of source code fragments in open source communities, extracts the core code associated with vulnerabilities through program slicing techniques, and constructs a BiLSTM model for malicious code detection. The text features of posts are also constructed using TF-IDF to train OVO SVMs models for vulnerability code type identification.