Differentially Private Bipartite Consensus for Multi-Agent Systems via Event-Triggered Communication With Edge-Based Additive Noise

Abstract

This article investigates the problem of event-triggered differential privacy bipartite consensus control in signed networks. Firstly, to address the dual challenges of initial state privacy leakage and communication efficiency in multi-agent systems, a distributed dynamic threshold event-triggered privacy-preserving algorithm is proposed. Secondly, the algorithm adopts an edge-based noise injection mechanism, independently adding noise to each communication channel at the event-triggered instant to defend against collusion attacks. Furthermore, a fully distributed control architecture is constructed, enabling each agent to make independent decisions based solely on its own state and local noisy information. Theoretical analysis demonstrates that the system converges in mean square, satisfies $\epsilon $ -differential privacy, and almost surely excludes Zeno behavior, while also providing a probabilistic lower bound on the event interval. Simulation results validate the superiority of the proposed algorithm in terms of privacy protection strength, convergence accuracy, and topological adaptability. Note to Practitioners—This article addresses the challenges of privacy and communication in multi-agent cooperative control. In practical deployment scenarios such as autonomous vehicles or distributed sensor networks, frequent data exchange not only consumes bandwidth but also exposes sensitive information. To address this issue, this article adopts a privacy-preserving mechanism that integrates dynamic threshold-based event triggering with edge-based noise injection. While reducing unnecessary transmissions, the proposed scheme prevents adversaries from inferring individual agents’ sensitive states from output information by flexibly designed noise on communication links. This approach provides a practical solution for practitioners designing resource-constrained, privacy-sensitive distributed systems, ensuring consistency accuracy. Simulation results demonstrate that the strategy achieves significant communication resource savings while maintaining privacy protection levels.