Differentially private Nash equilibrium seeking for networked aggregative games

TL;DR

This paper develops a distributed, privacy-preserving algorithm for Nash equilibrium seeking in networked aggregative games, ensuring differential privacy while accounting for communication constraints and convergence accuracy.

Contribution

It introduces a novel differentially private Nash equilibrium seeking method using perturbed average consensus in a fully distributed networked setting.

Findings

The proposed algorithm achieves $psilon$-differential privacy.

Convergence is proven under fixed and time-varying topologies.

Tradeoff between privacy level and convergence accuracy is analytically characterized.

Abstract

This paper considers the privacy-preserving Nash equilibrium seeking strategy design for a class of networked aggregative games, in which the players' objective functions are considered to be sensitive information to be protected. In particular, we consider that the networked game is free of central node and the aggregate information is not directly available to the players. As there is no central authority to provide the aggregate information required by each player to update their actions, a dynamic average consensus protocol is employed to estimate it. To protect the players' privacy, we perturb the transmitted information among the players by independent random noises drawn from Laplace distributions. By synthesizing the perturbed average consensus protocol with a gradient algorithm, distributed privacy-preserving Nash equilibrium seeking strategies are established for the aggregative games under both fixed and time-varying communication topologies. With explicit quantifications of the mean square errors, the convergence results of the proposed methods are presented. Moreover, it is analytically proven that the proposed algorithm is $\epsilon$-differentially private, where $\epsilon$ depends on the stepsize of the gradient algorithm and the scaling parameter of the random variables. The presented results indicate that there is a tradeoff between the convergence accuracy and the privacy level. Lastly, a numerical example is provided for the verification of the proposed methods.