Research on Dynamic Honeynet Technology Based on Machine Learning

In order to change the passive situation of the defender in the industrial control system security confrontation, honeypot honeynet as a typical deceptive defense technology has received wide attention. The static characteristics of traditional honeynet technology make it limited when dealing with complex network environments, for this reason, artificial intelligence technology was introduced to propose a threat trapping technology based on machine learning algorithms for industrial control systems. Aiming at the problems faced by the power industrial control system, such as the complexity of the type of anomalous traffic and the high dimensionality of the characteristics, the Random Forest algorithm is introduced to design the data detection module of the dynamic honeynet system, which improves the accuracy of the anomalous traffic identification. Meanwhile, the DQN algorithm of deep reinforcement learning is utilized to help the honeynet system make dynamic adjustment of strategies when facing different attack behaviors. Based on the typical network security dataset CIC-IDS-2017, the feasibility and effectiveness of the above design were verified. The experimental results show that the technique can effectively improve the network security protection capability of industrial control systems.