Industrial control network intrusion detection based on automatic optimization of hyperparameters

Summary: In view of the increasing number of cyber attacks on industrial control systems, it is difficult to determine the optimal hyperparameter group of existing deep learning intrusion detection models, soDue to the problem of low detection accuracy, a combination of hyperparameter optimization algorithm and stacked long and short-term memory is proposed.(SLSTM)The network-based industrial control system intrusion detection model realizes the automatic optimization of hyperparameters of the industrial control intrusion detection model based on deep learning, effectively improving the detection accuracy.First, use boundary-based synthetic minority oversampling technique to solve the problem of sample imbalance in the original data.Then, use the Bayesian optimization algorithm to search for the optimal hyperparameter group of the stacked long-term memory network.Finally, compare the proposed algorithm with other related algorithms on the industrial control network standard data set.Experimental results show that the proposed algorithm has the highest applicability in industrial control intrusion detection.

Keywords