LibDI: A Direction Identification Framework for Detecting Complex Reuse Relationships in Binaries
Siyuan Li, Chaopeng Dong, Yongpan Wang, Wenming Liu, Weijie Wang, Hong Li, Hongsong Zhu, Limin Sun · 2023
With the continuous evolution of software development, the increasing reuse and complexity of programs have become prominent. Detecting and analyzing reuse relationships in binary programs are crucial for software maintenance, performance optimization, and security enhancement. However, existing methods only identify whether binaries have reuse relationships without distinguishing the source binary of the reused code. This limitation hinders the detection of complex reuse relationships, such as nested reuse and pseudo-propagation reuse, and adversely affects the detection of 1-day vulnerability propagation relationships. To address this challenge, we propose LibDI, a Direction Identification framework specifically designed for detecting complex reuse relationships in C/C++ binaries. LibDI utilizes the Multi-level direction identification technique to accurately determine the direction of reuse between binaries, enabling the identification of complex reuse relationships. Experimental results demonstrate that LibDI achieves a recognition accuracy of 0.976 on public datasets, significantly outperforming existing methods. Moreover, LibDI successfully identifies complex reuse relationships in large-scale IoT firmware and effectively detects the propagation paths of vulnerabilities.