Nonlinear stealthy attacks on remote state estimation☆
Abstract
This paper investigates stealthy attacks on state estimation within linear time-invariant systems, focusing on scenarios where transmitted measurements through wireless networks are vulnerable to interception and manipulation by attackers. Their goal is to maximize estimation errors without triggering anomaly detectors. Moving beyond conventional linear attacks, this study introduces a comprehensive attack model that embraces nonlinear mappings. The evaluation of stealthiness hinges on several factors: the whiteness of residuals, the detector's detection rate, and the expectation of the detection index. Optimal attacks for both zero and nonzero detection rates are derived, incorporating various stealthiness constraints. As a general case, the optimal attack with any specified upper bound of detection rate can be designed, revealing that such attacks are predominantly nonlinear. Additionally, the paper explores the discontinuity of nonlinear mappings in these attacks, enhancing understanding of their complexity and implications. (c) 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
