A machine-learning approach for identification and mitigation of cyberattacks in networked process control systems

Abstract

The robustness of process control systems to cyberattacks has become an increasingly important topic, especially in light of the increased dependence of process operations on networked control system architectures wherein dedicated sensor-controller and controller-actuator links are replaced by real-time, wired or wireless, shared communication networks. To mitigate the risks of cyberattacks, we propose in this work an integrated model-based framework for the detection, estimation and mitigation of false data injection cyberattacks in nonlinear networked process systems. The framework leverages a classification-based neural network scheme for attack detection and estimation, coupled with a stability-based attack mitigation strategy based on a model-based controller design. An illustrative chemical process example is used to demonstrate the effectiveness of the proposed framework and assess its robustness with respect to possible attack detection and estimation errors.