Abstract
This paper investigates the man-in-the-middle (MITM) attack against cyber-physical systems (CPSs) under the random access protocol (RAP) scheduling, where an attacker intercepts and modifies the transmitted data and then forwards them on to degrade the system performance. The RAP schedules the sensing devices to avoid data collisions, where only one node is allowed to access the shared communication channel at each time instant. Hence, it makes the existing stealthy attacks invalid. To overcome the protocol-induced effects, a novel attack model utilizing only part of the measurements at each time instant is proposed, based on which the strictly stealthy and ∊-stealthy attacks are designed. For strictly stealthy attack, the Kullback-Leibler divergence (KLD)-based stealthy constraint is converted into a linear matrix inequality and then a semi-definite program problem is constructed to obtain the optimal attack parameters. In such case, the attack performance is optimal but limited. Furthermore, an ∊-stealthy attack is proposed to achieve higher attack performance, where the analytical attack parameters are obtained by solving an off-line convex optimization problem. Finally, simulations are provided to verify the results.