A Variable Structure Learning Automata Approach for the Feature Reduction Problem of Intrusion Detection Systems

A variable structure learning automata based method is proposed in this paper for solving the feature selection problem in designing the intrusion detection systems. The proposed method can explore the problem's search space using reward and penalty mechanism of learning automata. The target of proposed method is to increase the accuracy rate of the designed intrusion detection system by selecting the most significant features. The UNSW-NB15 intrusion detection dataset is employed for investigating the proposed method. The results of the designed experiments demonstrated the performance dominance of the proposed method  for most experiments in contrast with some other well-known methods.