Optimal sensor attacks in cyber-physical systems with Round-Robin protocol

Abstract

This paper investigates the stealthy attack problem in cyber-physical systems (CPSs) under the scheduling effects of the Round-Robin protocol (RRP), where the transmission order of the sensor nodes is modeled as a periodic function and only one node is allowed to access the shared communication channel at each time instant. The adoption of the RRP can efficiently avoid data collisions, while it makes the existing innovation-based stealthy attack strategies invalid. Thus, the aim is to design an attack strategy under the protocol-induced behaviors to maximize the attack effect and remain strictly stealthy simultaneously. Different from the existing results, where only single-sensor application scenarios were considered, multi-sensor ones with limited bandwidth are investigated in this work. A new attack model is proposed to overcome the protocol-introduced effects, which utilizes only part of the historical and the current innovations at each time instant. Based on the attack model, the stealthy condition is obtained and converted into the form of linear matrix inequality (LMI). Then, the error covariance under the attack and the RRP is derived to quantify the attack effect. Subsequently, the optimal attack strategy is obtained by solving a semi-definite programming (SDP) problem. Finally, simulations are performed to demonstrate the results.