A Binary Code Vulnerability Mining Method Based on Generative Adversarial Networks

Generative adversarial networks (GAN) is one of the most promising methods of unsupervised learning in complex distribution in recent years. Gan is widely used to generate data sets for data enhancement. However, the existing binary vulnerability mining methods can be divided into three ways: static analysis, dynamic analysis and dynamic static analysis. The research on the method of fundamentally expanding the data set to achieve vulnerability mining also has strong application value. Therefore, aiming at the problem of too few binary code vulnerability data sets, this paper proposes a binary code vulnerability mining model based on generation countermeasure network. In particular, the proposed system also combines automatic code generation technology, fuzzy testing and symbol execution technology to further optimize and train the generator and discriminator in the generation countermeasure network model to generate high-quality data sets. The experimental results show that, The binary code vulnerability mining model based on generative countermeasure network proposed in this paper can effectively solve the problem of too few data sets.