A Stripped Binary Code Vulnerability Detection Method Based on Sensitive Function Identification

With the increasing number of embedded devices, the demand for these devices is continually growing. Even in today's fast-paced technological era, many vulnerabilities still exist in embedded devices. Firmware vendors often release code packages in binary form due to reasons such as copyright and security concerns, and they typically remove symbol information upon release, complicating reverse analysis. Symbol information is fundamental to analyzing binary code in firmware, which is crucial for the accuracy, precision, and recall of vulnerability detection in binary code. The absence of symbol information, coupled with optimizations during compilation and the mixed arrangement of data and instructions, further complicates vulnerability detection in binary code. To address these issues, we propose a method for stripped binary code vulnerability detection based on sensitive function identification, which extracts attributes from the stripped binary code in the target firmware and uses the Euclidean distance metric algorithm to identify and label sensitive functions in the binary code, detecting five types of potential vulnerabilities. Experimental results show that vulnerability detection based on sensitive function identification is feasible and significantly improves accuracy, precision, and recall.