Development of Intrusion Detection in Industrial Control Systems Based On Deep Learning

Industrial control systems (ICSs) are essential and inseparable part of industrial infrastructures. Industrial control systems have long been designed and established isolated from the outside world; however, due to the needs for development and performance improvement, these industrial systems have been connected to other organization networks. Since security requirements and predictions have not been considered, ICSs are faced with new security threats. Therefore, cybersecurity in industrial control systems is of utmost importance due to severe economic, environmental, human and political consequences. Hence, the design of intrusion detection systems based on industrial control systems is also essential. In the present study, an accurate ICS scheme is developed based on the capabilities of deep neural networks (DNNs). In the proposed scheme, we try to detect the spatial space of packets exclusively by employing convolutional networks. Passing through the long short-term memory (LSTM) network, the time dependence between packets is used to diagnose abnormalities and attacks. Show that the proposed intrusion detection system outperforms other existing industrial intrusion detection systems in terms of accuracy. The high ability of the proposed scheme in dealing with unbalanced data sets is another exciting feature of the proposed scheme.