A Study on Deep Graph Neural Networks for Security Vulnerabilities Detection in Web Applications

Security becomes a top priority as the role of information technology and communication continues to grow, accompanied by the proliferation of web applications in every aspect of life. With every interactive application handling the information of thousands of customers, even a single vulnerability can lead to cascading effects and significant harm to individuals and organizations. The demand for tools and techniques to support vulnerability assessment and detection is increasing to mitigate the potential for such serious consequences. Furthermore, with the advent of data-driven approaches, deep learning to assist in vulnerability detection and software security assessment is gaining significant attention. In this paper, we investigate the ability to predict security vulnerabilities in source code using a combination of data mining and deep learning techniques. Specifically, we focus on improving the efficiency of detecting vulnerabilities in websites written in PHP by employing a combination of Deep Graph Convolutional Neural Network (DGCNN) and Code Property Graph (CPG). Furthermore, we propose a novel feature engineering technique tailored for code analysis and evaluate its performance compared to existing methods. Our results demonstrate the efficacy of our approach in enhancing vulnerability detection, offering a promising direction for future research and development in software security.