Privacy-Preserving Push-sum Average Consensus Algorithm over Directed Graph Via State Decomposition

Average consensus is the key basis of distributed collective behaviors of multi-agent systems. Almost all the existing average consensus algorithms require exact values of agents, under which the privacy of nodes is likely to be revealed to honest-but-curious neighbors. In this paper, we are concerned with the average consensus issue without loss of privacy of agents over a general directed network. A privacy-preserving push-sum algorithm is constructed for each agent based on state decomposition, where each agent sends its partial states instead of exact states to its neighbors. Such an algorithm not only guarantees the asymptotic average consensus but also preserves the initial value of each agent from disclosure. Finally, a numerical example is provided to verify the effectiveness of our algorithm.