Resilient distributed averaging: Adversary detection and topological insights

Abstract

In this paper, a distributed averaging algorithm in the presence of adversarial Byzantine agents is proposed. The algorithm is based on a resilient retrieval procedure, where all non-Byzantine nodes send their own initial values and retrieve those of other agents. We establish that the convergence of the proposed algorithm relies on strong robustness of the graph for locally bounded adversaries in a network with communication delays and asynchrony. A topology analysis in terms of time complexity and relation between connectivity metrics is also presented. Simulation results are provided to verify the effectiveness of the proposed algorithms under prescribed graph conditions.