Patch-based vulnerability discovery techniques

In today's software landscape, software architecture has become increasingly complex, with a wide range of functionalities, making software vulnerabilities almost unavoidable. Consequently, these vulnerabilities pose a significant risk to information systems security, directly impacting national strategic security, personal privacy, and property security. One solution to address this problem is to find and verify the exploitable vulnerabilities in software and take targeted measures to protect information system security as soon as possible. However, existing patch diffing techniques suffer from low accuracy in cross-architecture diffing and are rarely carried out at the binary level. To overcome this challenge, we propose a novel patch-based vulnerability mining technique that leverages patch diffing analysis to accurately locate vulnerable code and conduct effective vulnerability mining, making the discovery process more efficient. Specifically, we research a method for screening candidate pairs of matching functions based on the 3D-CFG algorithm, which identifies corresponding basic blocks within the program code, and a repair trace recognition method based on basic blocks. We also introduce a security patch identification approach based on semantic analysis, which assists in solving the problem of low accuracy in patch diffing. We conducted experiments on 1,000 function instances in real-world OpenSSL libraries, and the results demonstrate that our technique achieved an 85% accuracy rate. By addressing the limitations of existing patch diffing techniques and leveraging advanced algorithms, our proposed approach has the potential to significantly improve the effectiveness of vulnerability mining in software systems, thereby enhancing information system security