Localization and Detection toward Securing Networked Control Systems

Project: Research

View graph of relations

Description

This project initiates, from a systems and control perspective, an investigation into security issues of networked control systems with an orientation toward industrial control. Our goal is to develop a control-theoretic approach integrated with machine learning techniques, for the development of detection algorithms and defense strategies against cyber-attacks found in networked control systems. We propose a cohesive research plan consisting of four synergistic technical objectives, leading progressively to an integrated control-theoretic and machine-learning approach. First, we seek to develop a geometric control characterization and localization scheme for detectability of stealthy cyber-attacks. This unique scheme will enable us to identify and localize the structural vulnerabilities of a system where it may be vulnerable and the attack may be stealthy. Second, we attempt to develop detection algorithms and defense strategies for large-scale interconnected multi-agent systems (MASs) controlled by distributed controllers. The security of MASs poses more difficult and also more realistic a challenge, since distributed actuators and sensors expose them to attacks of a higher cardinality. Because of the scale and complexity of such systems, the detection algorithms must be scalable. We investigate fundamental requirements on sensor placements for detection and prevention of attacks, in both cardinality and geometry. In tight synergy with the above objectives, we shall also develop machine learning techniques to reinforce the control-theoretic approach, with the goal of developing integrated model-based control-theoretic and data-driven machine-learning algorithms for detection and estimation. Finally, the benchmark, industry-standard TennesseeEastman process will be employed to guide and validate our theoretical development.Industrial networked control systems are omnipresent in civil and industrial infrastructures, including, e.g., power grids, chemical processes, intelligent manufacturing, and hence are of enormous industrial, environmental, and societal impact. Unfortunately, networked control systems are also vulnerable to cyber-attacks. The project, while relevant to general cyber-physical system security problems, is motivated by and targets at challenges facing today’s industrial control systems security, and it seeks to provide answers and solutions addressing the challenges. We hold the viewpoint that the understanding generated from this project will contribute toward the advance and application of security studies for industrial networked control systems. The technical approach is built on the PI’s research results on networked control, multiagent systems and ongoing studies on cyber-physical security. The PI’s preliminary investigations show that the proposed approach is potentially feasible and may deliver results of significance. 

Detail(s)

Project number9043136
Grant typeGRF
StatusActive
Effective start/end date1/01/22 → …