Optimal Man-In-The-Middle Stealth Attack

In this paper we propose an optimal Man-In-The-Middle attack strategy to maliciously manipulate information transmitted from the field to a centralized control unit. The aim of the attacker is to significantly deviate the system’s behavior from its nominal trajectory and, at the same time, avoid that thew attack can be recognized. Specifically, we consider a scenario where an attacker is able to intercept and manipulate data transmitted from a remote sensor to a centralized state estimator equipped with a bad data detector. As shown in the paper, under the assumption of perfect information, the attacker, by solving a non-concave maximization problem, can cause a large discrepancy between the actual state and the estimated one without being discovered. By developing local optimality conditions for the problem at hand, we are able to design an algorithm to find an approximated solution. The paper is concluded by an example showing the best strategy for the attacker.