BarterCast

Abstract

Peer-to-peer (p2p ) file sharing systems rely on the sharing of resources by large fractions of partic- ipants. Sharing-ratio enforcement provides a very strong incentive for peers to contribute, leading to a higher performance for all peers in the system. One way of implementing this is by banning peers that do not share enough, determined by their past behavior (i.e., reputation). Various reputation mechanisms have been designed to facilitate this, but they are centralized or not feasible in practice. In this paper, we present a secure, fully distributed mechanism for reputation management and its integration with BitTorrent. The resulting system enforces a long-term balanced sharing-ratio for all peers in a BitTor- rent file sharing network. In our system, up- and download statistics are spread among peers and used to compute a subjective reputation for each peer. We apply the maxflow algorithm to limit the effect of peers that spread false information. We present simulations that demonstrate the system’s accuracy and effectiveness. The resulting system overcomes the long-standing problem of BitTorrent’s lack of seeding incentives, without any need for centralized administration, authority, or technology.