Privacy‐Preserving Achieving Average Consensus for Multi‐Agent Systems via Partial Information Transmission Against Curious Nodes Collusion

ABSTRACT

This paper investigates the privacy‐preserving average consensus problem in multi‐agent systems against collusive attacks, with the core objective of achieving accurate average consensus while protecting the privacy of agents' initial states. Existing methods are mostly limited to scenarios where curious nodes conduct isolated information theft, and are thus difficult to cope with the more practically threatening behavior of multi‐node collusive information theft. Based on the partial information flow algorithm, this paper designs anti‐collusion schemes from the node and communication link perspectives, respectively, and achieves resource savings by generating subgraphs with simplified structures. Theoretical analysis demonstrates that the proposed method can satisfy both privacy preservation and average consensus requirements in the presence of colluding curious nodes, and numerical simulation results further validate the effectiveness of the algorithms.