A new intrusion detection model based on belief rule base with variable interval structure for industrial control system

Abstract Intrusion detection in industrial control systems is of great significance to maintain the normal operation of infrastructure. However, the current common intrusion detection models cannot effectively use the semi-quantitative information consisting of expert knowledge and quantitative data, and most of them lack interpretability. Belief Rule Base (BRB), as a hybrid system, can effectively utilize both expert knowledge and historical data. Therefore, this paper proposes a new intrusion detection model for industrial control systems based on a variable interval structure BRB (VI-BRB). Firstly, the interval structure is introduced into BRB. By changing the conjunctive into disjunction, the method of rule generation is changed, and the problem of combinational explosion is effectively solved. Secondly, in view of the difficulty of determining the model interval, a variable interval structure is proposed. By changing fixed intervals into variable intervals, the optimal model structure can be found to improve model performance. Thirdly, an improved enhanced whale optimization algorithm (E-WOA) is proposed to handle the complex constraints in the optimization process of BRB while maintaining the interpretability of the model as much as possible. Finally, by conducting experiments on a natural gas pipeline dataset, VI-BRB achieved an accuracy of 97.3%. The experimental results indicate that VI-BRB can maintain high interpretability while obtaining high accuracy.