Detecting adversaries in distributed estimation

This paper studies distributed parameter estimation in the presence of adversarial agents. We consider the Flag Raising Distributed Estimation (FRDE) algorithm, a consen-sus+innovations distributed algorithm for the non-compromised agents to simultaneously perform parameter estimation and detect the presence of adversaries. So long as the non-compromised agents form a connected network and are globally observable, then, we can show that the FRDE algorithm either leads the non-compromised agents to correctly estimate the parameter or detect the adversarial activity preventing correct estimation. We demonstrate the performance of the FRDE algorithm through numerical examples.