Detection of buffer overflow vulnerability in embedded development programs based on SVM algorithm
Chuanwen Xiong · 2025
The filtering process for buffer overflow vulnerabilities is reversed, resulting in call errors and loop copy out of bounds errors. There are issues with low coverage of overflow vulnerability state detection, sensitive information leakage or tampering, and high deviation of buffer boundary values. To investigate the detection method of buffer overflow vulnerability in embedded development programs based on SVM algorithm. Consider buffer utilization for embedded development program buffer data capture, perform reverse data stream completion and denoising based on least squares fitting algorithm, and remove error patterns; Based on vulnerability data entropy, fuzzy rules are used to extract vulnerability features of buffer data flow overflow, and compared with reverse data flow to reduce the probability of sensitive information leakage or tampering; Implement overflow vulnerability detection based on the optimal classification function of SVM algorithm. The experimental results show that the overflow vulnerability state detection coverage of this method is always above 90%, reducing the deviation of buffer boundary values to around 0.5 and effectively achieving defense.