Fighting peer-to-peer SPAM and decoys with object reputation

Abstract

Peer-to-peer filesharing is now commonplace and its traffic now dominates bandwidth consumption at many Internet peering points. Recent studies indicate that much of this filesharing activity involves corrupt and polluted files. This paper describes Credence, a new object-based reputation system, and shows how it can counteract content pollution in peer-to-peer filesharing networks. Credence allows honest peers to assess the authenticity of online content by securely tabulating and managing endorsements from other peers. We employ a novel voter correlation scheme to weigh the opinions of peers, which gives rise to favorable incentives and system dynamics. We present simulation results indicating that our system is scalable, efficient, and robust.