A Transformer-based Approach for Automatic Vulnerability Detection and Classification

Abstract

With the development of science and technology, the application of software is becoming more and more extensive, and people's dependence on software is also increasing. Among them, software vulnerability is the primary security risk of software and the main factor affecting the stability and effectiveness of software. Predicting the existence of loopholes in the program has become an urgent task to solve security and stability problems, so it has received extensive attention and attracted a large number of researchers. At the same time, the development of machine learning provides a new method for software vulnerability and defect prediction, and code vulnerability prediction based on machine learning has become a research hotspot. In the current research results, most of the deep learning model Bi-RNN is used to solve the problem of automatic detection of vulnerabilities. This article proposes a more effective method of automatic detection of vulnerabilities. This method is based on the deep learning model Transformer to improve the accuracy of detection, and at the same time Greatly improve the detection speed. Compared with the existing research results VulDeePecker, the accuracy rate is increased by 1.51%, and the detection speed is increased by 69%.