Design And Implementation Of Industrial Control Network Intrusion Detection System Based On Deep Learning

Industrial control system,as a collection of all kinds of system equipment,network and controller to control and assist industrial automation production process,has been widely used in energy,electricity,water conservancy,pipeline,steel manufacturing and other fields,and the national and social operation,people’s production and life have a close connection,become an important part of public infrastructure.If there are loopholes or unknown attacks in the industrial control system,it may pose a major threat to social security.How to ensure the safety of industrial control system has become one of the research directions of public security strategy.Intrusion Detection System is the key of network security system,which can monitor the network and warn the relevant personnel to take measures when abnormal occurs.However,in the face of attacks,the past intrusion detection system is prone to false positives and missing positives,and there are some problems such as low training accuracy and great influence of human experience on detection effect.To solve the problems,this essay studies the anomaly detection method of industrial control network based on deep learning and designs and implements the intrusion detection system.The main work is as follows:(1)An industrial control feature selection method based on H-SVM-RFE is proposed to mine the relationship between industrial control data features.For the feature subset that was deleted in the feature selection iteration process,H-SVM-RFE reevaluated the correlation between the feature and the current feature subset from the correlation perspective combined with ABC,and the features with low correlation and redundancy were included back into the current subset,and the iteration was carried out.The experimental results show that H-SVM-RFE has better performance than other feature selection methods.On the neural network,the accuracy rate is 90.787%,and the F1 score is 0.86474.(2)The intrusion detection method of industrial control network based on CNN-BiLSTM is proposed.This essay expounds the characteristics of industrial traffic data,discusses the method of multi-scale convolutional neural network used for spatial feature extraction,and puts forward the construction of longterm and short-term memory network combined with attention mechanism to extract industrial time features from the study of industrial traffic sequence time features.Combining the advantages of multiscale convolution and long-term and short-term memory networks,a CNN-BiLSTM intrusion detection model is designed and its performance is analyzed.Through comparative experiments,it is proved that the classification accuracy of the model on the training set is 1.417% and 4.676% higher than CNN and LSTM,respectively,and the classification accuracy of the model on the test set is 94.879%.(3)An industrial control network IDS is designed.The key content includes the system module division,the industrial control network traffic data feature selection module based on H-SVM-RFE,which provides data support for industrial control network detection;Through the design of CNN-BiLSTM and attention mechanism,the spatial and temporal characteristics of network traffic data are extracted,and the network traffic data detection and classification are completed,providing the core detection capability of intrusion detection system;Finally,provide visual Web application interface display and data storage services for the test results.