Using Sensitivity Analysis and Cross-Association for the Design of Intrusion Detection Systems in Industrial Cyber-Physical Systems

Abstract

The fourth industrial revolution, also known as Industry 4.0, brings many advantages including innovative applications and services, new technologies and advanced features, increased operational benefits, and reduced installation costs. However, this technological advancement also exposes several challenges pertaining to the development of cyber-physical industrial architectures, resilient communication systems, and secure data exchange. This paper develops a systematic methodology for designing intrusion detection systems (IDS) specially tailored to address the cyber and physical dimensions of these systems. The approach is aimed at reducing the number of monitored parameters by adopting a three-phase design strategy embracing sensitivity analysis, cross-association, and optimal IDS design. To this end, phase 1 embraces sensitivity analysis to identify sensitive variables to specific interventions (e.g., control signals and cyber attacks), phase 2 adopts the cross-association assessment to optimally structure the process variables in groups that are the most sensitive to groups of interventions, and finally, phase 3 optimally assigns the most sensitive process variables to IDS, while enforcing the IDS capacity limitations and redundancy requirements. Numerical results on a realistic vinyl acetate monomer process show that the approach can reduce the number of variables by 76.8%, thus reducing the complexity and the costs of the detection infrastructure.