Firmware Vulnerability Detection Algorithm Based on Matching Pattern-Specific Numerical Features With Structural Features

With the continuous improvement of Internet of Things technology, Internet of Things devices are gradually popularized in people’s lives and work, bringing convenience to people, but also there are many security risks. There are more and more types of attacks on IoT devices, and the security of their firmware has become a focus of attention. Aiming at firmware vulnerabilities in devices, a firmware vulnerability detection algorithm based on pattern-specific features and structural features is proposed in this study. The algorithm uses the two-stage method to filter and match the functions precisely, so as to find the functions matching the vulnerability functions. By reducing the local call graph from five layers to three layers, the algorithm operation and detection efficiency are improved, and the accurate matching method of weighted three-layer local call graph is implemented. The experimental results showed that the Top1 index value of the five-layer local call graph ranges from 81.99 to 90.19. The indexes of control flow chart and attribute control flow chart fluctuated greatly, ranging from 61.57 to 91.08 and 54.62 to 87.55, respectively. The Top1 index value of the weighted three-layer local call graph increased by 3.73%, and the average increased by 1.47%, indicating a significant improvement in the whole. It can be concluded that the firmware vulnerability detection algorithm based on the matching of pattern-specific numerical features and structural features can effectively find the function that actually matches the vulnerability function, and improve the efficiency of firmware vulnerability detection.