(419a) Keynote Talk-on the Role of Control System Design on Detecting Cyberattacks

The increasing reliance on advanced communications and networking technologies and the increasing complexity and sophistication of cyberattacks have rendered industrial control systems, such as process control systems, vulnerable to cyberattacks by malicious agents [1-2]. Cyberattacks on control systems aim to disrupt operations, destabilize processes, or cause safety incidents [1-3]. A cyberattack against a control system may manipulate or disrupt the data communicated over the links between the sensors and controller or the controller and actuators. These attacks compromise the control system integrity, causing it to operate based on falsified or inaccurate data. The traditional approach for addressing the threat of cyberattacks involves fortifying information technology (IT) systems to prevent attackers from gaining access to the control system network. As evident by the myriad of recent cyberattacks, IT-based approaches are insufficient in preventing cyberattacks, and operational technology-based approaches are needed.

Detecting the presence of a cyberattack targeting the control system is one necessary OT-based capability. Cyberattack detection methods detect the presence of an attack by identifying abnormal operating behavior from operational data [4-6]. Several methods have been proposed, including chi-squared, cumulative sum, and data-driven and machine learning-based approaches (see, for example, the reviews [4-6]). A vector-valued model-based cumulative sum detection method was proposed [7]. The potential performance degradation that a stealthy attack can induce while remaining undetected was characterized. A generalized chi-squared detector for linear systems with non-Gaussian noise was proposed [8]. Three detection-related concepts were presented, including randomized switching of the control system to help detect attacks, a residual-based detection scheme using the difference between the measured state and predicted state as the residual, and a redundant state estimation-based detection scheme. Several models of cyberattacks were considered in [10]. Neural network-based detector schemes were developed and trained based on extensive simulation studies of the attack-free and attacked process [10]. While there has been a degree of acknowledgment that the control system design plays a role in the ability or inability to detect cyberattacks, the connection between control system design and cyberattack detectability has not been rigorously established.

In the present work, the role of controller design on cyberattack detectability is analyzed. Specifically, false-data injection attacks, attacks that manipulate the data communicated over the control system communication links, are considered. Rigorous definitions related to attack detectability are developed for a general class of detection schemes that encompass several commonly employed detection schemes. An analysis of attack detectability reveals that the control design impacts the ability to detect cyberattacks. These findings are verified using a chemical process where several detection schemes are considered under different control system designs. The results demonstrate that an attack may be detected or undetected depending on the control system design.

References

1. Hemsley, K., and Fisher, R., History of industrial control system cyberattacks, Idaho National Laboratory, Report number: INL/CON-18-44111, 2, 1-37 (2018).

2. Kayan, H., Nunes, M., Rana, O., Burnap, P., and Perera, C., Cybersecurity of industrial cyber-physical systems: A review, ACM Computing Surveys, (in press).

3. Ginter, A., The top 20 cyberattacks on industrial control systems, Technical Report, Version 1.1, Waterfall Security Solutions (2018).

4. Giraldo, J., Urbina, D., Cardenas, A., Valente, J., Faisal, M., Ruths, J., Tippenhauer, N. O., Sandberg, H., and Candell, R., A survey of physics-based attack detection in cyber-physical systems, ACM Computing Surveys, 51, 1-36 (2018).

5. Ding, D., Han, Q., Xiang, Y., Ge, X., and Zhang, X.. A survey on security control and attack detection for industrial cyber-physical systems, Neurocomputing, 275, 1674-1683 (2018).

6. Tan, S., J. M. Guerrero, P. Xie, R. Han, and J. C. Vasquez. Brief survey on attack detection methods for cyber-physical systems. IEEE Systems Journal, 14:5329-5339, 2020.

7. Murguia, C., Ruths, J., CUSUM and chi-squared attack detection of compromised sensors, Proceedings of the 2016 IEEE Conference on Control Applications, 474-480 (2016).

8. Hashemi, N., and Ruths, J., Generalized chi-squared detector for LTI systems with non-Gaussian noise, Proceedings of the 2019 American Control Conference, 404-410 (2019).

9.Oyama, H. and Durand, H., Integrated cyberattack detection and resilient control strategies using Lyapunov-based economic model predictive control, AIChE Journal, 66, e17084 (2020).

10. Chen, S., Wu, Z., and Christofides., P. D., Cyber-attack detection and resilient operation of nonlinear processes under economic model predictive control, Computers & Chemical Engineering, 136, 106806 (2020).