Research on a New Vulnerability Risk Detection Method Based on Bidirectional Constructed Hybrid Attribute Graphs

Web application network security incidents such as tampering with web pages, denial of service, stealing and destroying data are seriously threatening the normal operation of power websites. Aiming at the low detection rate of Web vulnerability detection methods, a source code vulnerability detection model based on bidirectional construction mixed attribute graph is proposed. The method converts source code into a mixed attribute graph, which is used to capture the syntax and semantic information of the code in the model. By analyzing the different structural characteristics of the vulnerability sample and the correct sample, the vulnerability detection of source code is transformed into a graph classification problem. The model was tested on a public data set, and the results show that the F1 score of the model tends to be stable after 5 iterations. When the number of convolutional pooling layers is 3, the detection effect is the best, and the F1 value reaches the maximum. According to the model, 7 of the 11 test cases were fully matched and 2 were partially matched, and the detection rate reached 81.81%. The detection effect of the proposed model is more accurate and stable, and better vulnerability detection results are obtained, which can realize all-round protection and real-time detection alarm of the security events of the power website and its web business system, and improve the emergency handling ability of the power system for web security events.