Joint Measurement and Channel Design of a Malicious Sensor in Distributed Estimation Based on Maximum Disturbance in a Sensor Network

Abstract

Secure distributed estimation algorithms aim to be resilient against adversaries in a network. By deploying a single attacker with sufficiently large attack vectors in the network, the adversary can significantly degrade the performance of the estimator. Large attack vectors enhance the chance of attack detection. This letter aims to optimally design the measurement and channel attack vectors of a single attacker to maximally deviate the performance of the distributed estimation algorithm based on the maximum disturbance. A suboptimal joint measurement and channel attack design are provided using a Lagrange multipliers' method, in which the Lagrange multipliers are arbitrary and not obtained optimally. Subsequently, a suboptimal design for measurement-only and channel-only attacks is presented, with Lagrange multipliers derived mathematically. In fact, the false data injection (FDI) of a sensor has a profound effect on the performance of the distributed estimation in a sensor network. So, the action of even a single malicious sensor with deliberate attack design can degrade the true performance of the entire network. Simulation results demonstrate that these attack designs algorithms are more effective than random attack designs.