Innovation-based Linear Stealthy Attacks on Kitanidis Filtering

This paper investigates stealthy attacks targeting remote state estimation processes that utilize the Kitanidis filter, focusing on identifying optimal strategies to compromise system performance in linear time-varying systems with unknown inputs. The research aims to craft attacks that maximize degradation in cyber-physical systems’ performance while remaining undetectable by false data detectors. The key part of this study is the analysis of innovation-based linear attacks, which manipulate the estimation error covariance via exploiting the characteristics of the Kitanidis filter. By tracking the dynamic evolution of the estimation error covariance, the study identifies an optimal stealthy attack strategy that drives the system into its worst-case performance scenario. This not only highlights vulnerabilities in state estimation mechanisms but also offers insights into the attack impact on critical system stability and accuracy. The theoretical findings are validated through a detailed numerical simulation, which demonstrates the severe disruptions that such stealthy attacks can introduce.