DSHGT: Dual-Supervisors Heterogeneous Graph Transformer -- A pioneer study of using heterogeneous graph learning for detecting software vulnerabilities

TL;DR

This paper investigates vulnerability detection in software by leveraging heterogeneous graph learning on Code Property Graphs with a Dual-Supervisors Heterogeneous Graph Transformer (DSHGT). By combining a graph representation that captures rich intra-code relations with a dual supervision signal from code comments, the approach achieves state-of-the-art results on SARD and demonstrates cross-language transferability from C/C++ to Java and PHP. Experiments on real-world projects FFmpeg, QEMU, and Wireshark show consistent improvements in accuracy, precision, recall, and F1, highlighting practical viability. The work points to future directions like integrating raw code tokens and refining supervisory signals to further enhance vulnerability detection in diverse software ecosystems.

Abstract

Vulnerability detection is a critical problem in software security and attracts growing attention both from academia and industry. Traditionally, software security is safeguarded by designated rule-based detectors that heavily rely on empirical expertise, requiring tremendous effort from software experts to generate rule repositories for large code corpus. Recent advances in deep learning, especially Graph Neural Networks (GNN), have uncovered the feasibility of automatic detection of a wide range of software vulnerabilities. However, prior learning-based works only break programs down into a sequence of word tokens for extracting contextual features of codes, or apply GNN largely on homogeneous graph representation (e.g., AST) without discerning complex types of underlying program entities (e.g., methods, variables). In this work, we are one of the first to explore heterogeneous graph representation in the form of Code Property Graph and adapt a well-known heterogeneous graph network with a dual-supervisor structure for the corresponding graph learning task. Using the prototype built, we have conducted extensive experiments on both synthetic datasets and real-world projects. Compared with the state-of-the-art baselines, the results demonstrate promising effectiveness in this research direction in terms of vulnerability detection performance (average F1 improvements over 10\% in real-world projects) and transferability from C/C++ to other programming languages (average F1 improvements over 11%).

Figures (12)

• Figure 1: Sample code and the generated CPG.
• Figure 2: Graph Construction
• Figure 3: DSHGT Learning
• Figure 4: Code sample of QEMU and FFmpeg, along with corresponding generated dot file.
• Figure 5: Code sample of Wireshark and manifest.sarif file, along with the corresponding dot file.