(14d) Post Cyber-Attack Secure State Reconstruction for Nonlinear Processes Using Machine Learning | AIChE

(14d) Post Cyber-Attack Secure State Reconstruction for Nonlinear Processes Using Machine Learning


Chen, S. - Presenter, University of California, Los Angeles
Wu, Z., University of California Los Angeles
Christofides, P., University of California, Los Angeles
Rincon, D., University of California, Los Angeles
In recent years, several works have been dedicated to investigating various aspects of securing complex chemical plants, e.g., by ensuring the networking security amongst cyber devices, the prompt detection of faults and attacks, as well as accurate estimation of states using corrupted state measurements [1,3]. As it is important to develop accurate detectors to promptly report the intrusion of a cyber-attack as well as building robust frameworks to mitigate the impact of cyber-attacks before the detector is activated, it is equally important to have recuperating measures in place to maintain controllability of the system in the absence of reliable sensors.

This work investigates state-reconstruction strategies to maintain the controllability of the system effectively following the detection of cyber-attacks on sensor measurements. Considering a general class of nonlinear systems of which the sensor measurements are vulnerable to malicious cyber-attacks, there have been previous works on robust control frameworks that effectively maintain the stability of the process in the presence of cyber-attacks, within which the detection and differentiation of various types of cyber-attacks have been successfully carried out by machine-learning-based detection algorithms [2,4]. In this work, we further explore recuperation measures as a response plan after the detection of cyber-attacks to minimize or eliminate its impact. A machine-learning-based state reconstruction approach is presented to provide estimated state measurements based on falsified state measurements, and it ensures stable operation of the process before reliable sensor measurements can be re-installed.

[1] Ao, W., Song, Y., Wen, C., 2016. Adaptive cyber-physical system attack detection and reconstruc-tion with application to power systems. IET Control Theory & Applications 10, 1458–1468.

[2] Chen, S., Wu, Z., Christofides, P.D., 2020a. Cyber-attack detection and resilient operation of non-linear processes under lyapunov-based economic model predictive control. Computers & ChemicalEngineering 136, 106806.

[3] Hu, Q., Fooladivanda, D., Chang, Y.H., Tomlin, C.J., 2017. Secure state estimation for nonlinearpower systems under cyber attacks, in: Proceedings of the American Control Conference, Seattle,Washington. pp. 2779–2784.

[4] Wu, Z., Albalawi, F., Zhang, J., Zhang, Z., Durand, H., Christofides, P.D., 2018. Detecting andhandling cyber-attacks in model predictive control of chemical processes. Mathematics 6, 173, 22pages.