Design and Implementation of Private Industrial Control Protocol Vulnerability Mining System Based on Deep Learning

Design and Implementation of Private Industrial Control Protocol Vulnerability Mining System Based on Deep Learning

연구 분야: Strategies

학회: 2024 8th Asian Conference on Artificial Intelligence Technology (ACAIT)

As the open environment greatly enhances the flexibility and efficiency of industrial control systems, it also makes security issues increasingly prominent. Once a problem occurs, it may have negative impacts on people's daily life, society, and even national security. Therefore, the security issues of industrial control systems are urgent to be solved. Industrial control protocols, as the bridge connecting every components of the industrial control system, are undoubtedly of great importance. Some industrial control protocols have security vulnerabilities in their initial design or implementation, These vulnerabilities are one of the reasons for security incidents in industrial control systems. Therefore, it is necessary to find vulnerabilities in the protocols, which can discover deficiencies in protocols before incidents occur, and promote the improvement of the protocols. In this paper, we implement a field boundary division algorithm based on information theory to divide protocol fields. We train SeqGAN model and use it to generate test cases. We design a monitoring feedback approach to adjust the fuzzing mutation values. During the experimental process, a new vulnerability was discovered.