A unifying approach for the identification of application-driven stealthy attacks on mobile CPS

Cyber-Physical Systems (CPS) employing mobile nodes rely on wireless communication in many critical applications. Due to interference and intentional jamming by adversaries, mobile nodes may fail to communicate with each other causing severe performance consequences. In this paper, we present a unifying approach for identifying attacks that target Mobile CPS applications. The attack policies are obtained as solutions to Markov Decision Process (MDP) problems, in which a decision to interfere with a signal on a given link is based on the current state of the system. Through applying approximate policy iteration methods, efficient attack policies that only interfere with a selective set of signals between the mobile nodes are derived to maximize damage while minimizing exposure and detection. The proposed approach is instantiated on pheromone-based coordination methods that are used in reconnaissance, surveillance, and search missions in military operations. The identified attack policies are shown to be more potent than other attack policies, including myopic, heuristic, and Denial of Service (DoS) policies.