(14c) Relating State Estimation, Data-Driven Modeling, and Computing to Control System Cybersecurity

Cybersecurity of process control systems is receiving increasing interest in the process systems engineering community due to the potential for cyberattacks to impact both safety and profits in the chemical and pharmaceutical industry. An important question is defining the conditions under which it can be guaranteed that even if a cyberattack is performed on an aspect of a control loop (e.g., by falsifying the sensor measurements provided to a controller, or by falsifying a signal received by an actuator), the process does not experience closed-loop stability issues. Directions which look like they may have promise for handling this issue include cyberattack detection methods using neural networks (e.g., [1]), or methods which attempt to reconstruct the state when only a limited number of sensor measurements are attacked in a linear system [2]. However, methods for guaranteeing that closed-loop stability is not lost under a cyberattack on a nonlinear system are still needed.

In this work, we explore several concepts for preventing cyberattack success for nonlinear control systems. First, we explore how state estimation for nonlinear systems can be utilized to develop mechanisms for preventing closed-loop stability issues when a limited number of sensors is attacked. In addition, we develop a methodology for probing for cyberattacks using Lyapunov-based economic model predictive control (LEMPC) [3] that randomly develops constraints based on state-space regions in which the manner in which a Lyapunov function changes over the next sampling period is known in the absence of an attack, so that if the expected trend is not observed over the next sampling period, a cyberattack may be occurring. We also explore how data-driven models used in optimization-based control designs could be impacted by falsified sensor measurements, and how such modeling errors could impact closed-loop stability. Finally, we explore how the computing and communication strategies with optimization-based control may impact cybersecurity considerations.

[1] Wu, Z., F. Albalawi, J. Zhang, Z. Zhang, H. Durand and P. D. Christofides, "Detecting and Handling Cyber-Attacks in Model Predictive Control of Chemical Processes," Mathematics, 6, 173, 22 pages, 2018.

[2] Fawzi, H., P. Tabuada and S. Diggavi. “Secure estimation and control for cyber-physical systems under adversarial attacks.” IEEE Transactions on Automatic Control. 59, 1454-1467, 2014.

[3] M. Heidarinejad, J. Liu and P. D. Christofides. Economic model predictive control of nonlinear process systems using Lyapunov techniques, AIChE Journal, 58:855-870, 2012.