Differentiated Output-Based Privacy-Preserving Average Consensus

Abstract

This letter investigates the differentiated output-based privacy-preserving average consensus problem over digraphs. A new stochastic obfuscation algorithm is proposed to achieve better privacy-preserving effect. When the output messages for at least one out-neighbour are not leaked, the algorithm can be designed to achieve any pre-given consensus accuracy and privacy-preserving level simultaneously by properly selecting the private weights. Even if all the output messages are leaked, the algorithm can still ensure that each agent’s initial state is protected to a certain extent. The mean square convergence of the algorithm is proved. The efficiency of the algorithm is verified by a numerical example.