Rational information exchange model: a new optimization approach for equilibrium computing

Abstract

This paper presents equilibrium conditions of multiparty information exchange problems. Pieces of information are exchanged among parties in a fair, peer-to-peer, private and simultaneous manner. The idea is so that at the end each party can construct secrets from what it receives. These secrets can either be false or true, which yield negative or positive payoffs. By formalising, we demonstrate that although cheating is a dominant strategy, there might be other less trivial equilibria in which parties are reluctant to information exchange. We devise a method for computing generalized equilibria, based on solving MLIP models. The necessary and sufficient conditions for the equilibrium existence can be tested by the MLIP model. We present case studies that illustrate our model features, essentially with multiple equilibria and with more than five players. The running time is exponential in the number of players and polynomial in the number of information pieces.