Distributed differentially private average consensus for multi-agent networks by additive functional Laplace noise.

Average consensus is a very useful consensus algorithm for distributed cooperative control and computing, where all the agents in the network communicate with its neighbor and reach the average of the initial states of all agents. The main secure defect of average consensus is that the initial state of agent can be inferred by using the state information sequence of the agent, which leads to the information disclosure. To handle this problem, in this paper, based on functional Laplace noise and differential privacy scheme, a novel differentially private average consensus algorithm is proposed to preserve the privacy of the state of each agent in the whole process of consensus computation. We develop detailed consensus analysis of our algorithm, including convergence rate and the consensus condition for network agents. Moreover, the privacy-preserving analysis is also given, which indicates that privacy of the states of all agents is guaranteed to preserve. Finally, a numerical experiment is used to demonstrate the effectiveness of our algorithm. [ABSTRACT FROM AUTHOR]