(245a) Control Techniques for Handling Sensor and Actuator Cyberattacks on Evolving Nonlinear Process Systems

Cyberattacks on control systems are a concern due to the possible catastrophic consequences of a successful attack. Despite various efforts in recent years to develop techniques for detecting attacks via machine learning [3] or state prediction [4], or handling attacks via other techniques (e.g., signal injection [5]), many open challenges with respect to control system cybersecurity remain. One of these challenges is that cyberattacks might be identified by evaluating deviations from “normal” process behavior. However, as the process dynamics change over time for a process, it could be that metrics designed to detect deviations from “normalcy” begin to flag dynamics changes as attacks. This concept motivated our work in [6], in which we extended two detection strategies from [7] that are integrated with the use of an optimization-based control strategy termed Lyapunov-based economic model predictive control (LEMPC) [1, 2] to the case where process dynamics change over time. Specifically, when there are no attacks or changes in the process dynamics, the closed-loop state is maintained within a characterizable region of operation at all times, and a threshold on either the difference between state predictions and measurements or between different state estimates is not breached. After either the closed-loop state leaves the characterizable region of operation or the difference between the predictions/measurements or estimates is breached, the threshold on the predictions/measurements or estimates is modified and model re-identification is triggered after a defined number of sampling periods. This strategy is able to theoretically guarantee that the closed-loop state is maintained within a characterizable operating region for at least one sampling period after an undetected attack or until the time the model re-identification occurs when there is only a model change and no attack.

Despite the elegance of the theories developed in [6], an important question to address is how such a strategy might be implemented practically. Specifically, the theoretical results in [6] introduce a number of theoretical expressions which must be satisfied, including, for example, various parameters related to the operating region and plant/model mismatch. Through a series of simulation examples involving a continuous stirred tank reactor and focused on the state estimation-based attack detection strategy, we elucidate some of the difficulties in designing resilient control strategies without rigorously determining the parameters of the control and detection strategies using theory but instead attempting to utilize simulation. We also discuss, via simulation, strategies for attempting to thwart specific types of attacks on the actuators via injecting control signals at intervals, as well as the role of plant/model mismatch introduced via data-driven models and numerical error on control actions when state measurement cyberattacks occur. Finally, in light of these control/detection concepts, we discuss concepts related to cyberattack discoverability for nonlinear systems.

References:

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

[2] Ellis, M., Durand, H., & Christofides, P. D. (2014). A tutorial review of economic model predictive control methods. Journal of Process Control, 24(8), 1156-1178.

[3] Wu, Z., Albalawi, F., Zhang, J., Zhang, Z., Durand, H., & Christofides, P. D. (2018). Detecting and handling cyber-attacks in model predictive control of chemical processes. Mathematics, 6(10), 173.

[4] Cárdenas, A. A., Amin, S., Lin, Z. S., Huang, Y. L., Huang, C. Y., & Sastry, S. (2011, March). Attacks against process control systems: risk assessment, detection, and response. In Proceedings of the 6th ACM Symposium on Information, Computer and Communications Security, 355-366.

[5] Satchidanandan, B., & Kumar, P. R. (2016). Dynamic watermarking: Active defense of networked cyber–physical systems. Proceedings of the IEEE, 105(2), 219-240.

[6] Rangan, K. K., Oyama, H., & Durand, H. (2021). Integrated cyberattack detection and handling for nonlinear systems with evolving process dynamics under Lyapunov-based economic model predictive control. Chemical Engineering Research and Design.

[7] Oyama, H., & Durand, H. (2020). Integrated cyberattack detection and resilient control strategies using Lyapunov‐based economic model predictive control. AIChE Journal, 66(12), e17084.