A game-theoretic approach to coding for information networks

Abstract

Since its inception, network coding literature has, for the most part, assumed cooperation among users. In unicast applications where users have no inherent interest in providing (or concealing) their information to (or from) any destinations except for their unique one, this assumption must be reconsidered. In this paper, we examine the impact of selfish users on coding strategies by formulating network coding games, in which users strategies are their encoding/decoding schemes (including encoding functions, block length, rate, etc). Through the use of examples, we show that the rational outcomes of such network coding games are dependent on the particular network coding scheme implemented at intermediate nodes in the network. More specifically, we construct examples that show how careful construction of network coding schemes at intermediate nodes in the network can guarantee that efficient coding solutions will emerge as a rational outcome of the game, even when users are allowed complete freedom in choosing their coding schemes.