AN APPROACH TO INTRUSION DETECTION FOR INDUSTRIAL CONTROL SYSTEM NETWORK BASED ON 1DCNN AND BiSRU

An intrusion detection method is proposed based on one-dimensional convolutional neural network (1DCNN) and bidirectional simple recurrent unit (BiSRU) to solve the problems of sample class imbalance and insufficient feature extraction in intrusion detection of industrial control system (ICS) network. The training samples were optimized by the synthetic minority oversampling technique in this method. The 1DCNN and BiSRU were used to extract the sample space features and contextual timing semantic information respectively. And this method performed sample multi-classification by fully connected layers. Simulation results show that the comprehensive performance of this method is much better than other algorithms, and it can effectively identify intrusion behavior of the ICS network.