Innovation-based stealthy attack against distributed state estimation over sensor networks

This paper presents a new design of an innovation-based stealthy attack strategy against distributed state estimation over a sensor network. In the absence of network attack, an optimal distributed minimum mean-square error (MMSE) estimator is developed by fusing the interaction measurements from neighboring nodes in the sensor network. Also, the boundedness of distributed estimation covariance is discussed over a regionally observable sensor network, which weakens the requirement for local observability of each sensor. Then, a stealthy attack framework embedded with an adjustable parameter is proposed, under which the attack strategy is to maximize the distributed estimation covariance. Sufficient conditions on the boundedness of the compromised covariance are derived, and the tradeoff between attack stealthiness and attack effects is determined. Finally, numerical examples are shown to verify the developed techniques. (c) 2023 Elsevier Ltd. All rights reserved.

Automated summary

This paper presents a new design of an innovation-based stealthy attack strategy against distributed state estimation over a sensor network. The optimal distributed MMSE estimator is developed by fusing interaction measurements from neighboring nodes in the absence of network attack. The tradeoff between attack stealthiness and attack effects is determined by proposing a stealthy attack framework embedded with an adjustable parameter.