Software vulnerability detection method and device based on graph convolution network

Abstract

The invention belongs to the technical field of computer security, and particularly relates to a software vulnerability detection method and device based on a graph convolution network, and the methodcomprises the steps: collecting a code sample data set; for the code sample data set, based on the abstract syntax tree and the control flow graph, extracting a code feature semantic expression formin code behavior units; and performing classification detection on the extracted code feature semantic expression form through the trained and learned graph convolutional network, and obtaining software vulnerabilities in the code according to a classification detection result. According to the invention, through extracting the abstract syntax tree, the control flow and the data flow information of the code, a code feature semantic expression mode taking a code as a unit is formed. The possibility that the vulnerability exists in the code is analyzed by using the graph convolution network in deep learning and the characteristics analyzed by the data flow extracted based on the abstract syntax tree, so that the vulnerability detection efficiency and accuracy are effectively improved, the software code reliability is ensured, the software quality is improved, and the method has important significance for software security application.

Description

technical field [0001] The invention belongs to the technical field of computer security, and in particular relates to a software vulnerability detection method and device based on a graph convolution network. Background technique [0002] Now that software technology has penetrated into all aspects of life, the security of these systems themselves is becoming more and more important. However, various loopholes in the software seriously threaten the security of the system. A total of 27,583 vulnerabilities were disclosed between 2015 and 2017, with an average of 9,194 vulnerabilities discovered per year. The number of vulnerabilities discovered has increased by more than 70% compared to an average of 5,400 vulnerabilities being disclosed per year between 2006 and 2014. There are a small number of uncovered vulnerabilities known as 0days that are favored by attackers. The number of zero-day vulnerabilities discovered in 2015 increased dramatically, averaging about one per ...

AI Technical Summary

Problems solved by technology

These methods work well in a theoretical environment, but in real scenarios, due to the

Images