Deep Learning Techniques to Detect DoS Attacks on Industrial Control Systems: A Systematic Literature Review

Abstract

Cyber Physical Systems (CPS) security is crucial demand within industrial fields. The deployment of these systems within critical infrastructure is increasing day by day. CPS applications include smart grid, Industrial Control Systems (ICS), Aerial Systems and Intelligent Transportation Systems (ITS). The complexity, heterogeneity, and diversity evolved with these CPS systems. In addition, the inter-connectivity of these systems over cyberspace has increased their attack surface. This research paper provides a survey on deep learning detection techniques for the Denial of Service (DoS) attack, which is considered the most critical and major attack on CPS. Moreover, the survey study demonstrates the most used deep learning techniques in the research articles of traditional IT networks and ICS networks. It also explains their used datasets as training sources and their most common evaluation matrix that is used to benchmark their performance against each other. In addition, the research gaps that are related to classifier efficiency are identified, while considering modern datasets related to ICS protocols. Moreover, consider the actual cyberspace attack traffic collected from passive monitoring sensors. This would resolve the need for using less features provided over outdated and publicly available dataset.