VulScan: A Vulnerability Detection Model Based on Deep Learning

Ensuring software security requires effective identification of vulnerabilities in the source code, as they can make software systems susceptible to attacks. In recent years, deep learning techniques have emerged as promising approaches for vulnerability detection in source code. This paper introduces VulScan, a model that combines Graph Convolutional Network (GCN) and Bidirectional Long Short-Term Memory (BiLSTM) to identify vulnerabilities. By representing the source code as an Abstract Syntax Tree (AST) and applying GCN to extract structural features, VulScan captures both the graph-based relationships and sequential information using BiLSTM. Comparative evaluations against baseline techniques commonly used in vulnerability detection demonstrate VulScan’s superior performance, leading to improved detection accuracy.