I. Introduction
Recently, cyber-physical systems (CPS) have drawn much research interest within the discrete-event systems and formal methods communities [1]–[9]. Due to the prevalence of cyber-physical systems in modern society and the catastrophic consequences that could happen after these systems get attacked [10], [11], it is important to design resilient control mechanisms against adversarial attacks. The first step towards this goal is to understand when adversarial attacks can succeed.