R-Chain: A Self-Maintained Reputation Management System in P2P Networks

Trust management is a critical component in P2P file sharing systems due to the free riding and security problem. In this paper, we propose a lightweight reputation management system, R-Chain, where each peer maintains its own transaction history as the reputation. In R-Chain, each transaction is supervised by a few randomly chosen witnesses and each peer keeps a record for every transaction it has participated in. To ensure the integrity, transaction records are signed by the witnesses and organized as a chronicle chain. With very limited data stored on the network, no peer can modify or discard these transaction records without being detected. Compared with other reputation systems, our design is more light-weighted in terms of reputation maintenance and retrieval. It is scalable since each transaction only involves a constant number of peers. By recording the raw transaction history, R-Chain can also be integrated with different trust models to prevent various attacks. 1 INTRODUCTION With the increase in network bandwidth, storage space and computing power of modern personal computers, information sharing systems using Peer-to-Peer (P2P) techniques have gained tremendous attention in both academy and industry. P2P systems like Gnutella and KaZaA are used by a large number of users everyday, albeit for not-so-legal reasons. P2P networks however hold much promise for sharing and collaborations between ordinary users without the need for controlling servers and corporate interests. While P2P networks provide an open and anonymous environment for information sharing, they are also vulnerable to various attacks, such as distributing tampered files or even spreading virus like VBS.Gnutella [1]. Due to the opaque-identification that each peer can create, change, discard its own identity, malicious peers cannot be identified and published for their misbehavior. Therefore, choosing a trustworthy peer for each transaction is very important. In addition, trust management can also be used to address the prevalent " free-riding " problem in P2P networks [2]. One key problem in a trust management system is the management of reputation data. Since it is stored and maintained on highly dynamic environment where unreliable peers keep joining and leaving frequently [2], it is critical to ensure that such data cannot be changed or lost. Two different methods have been proposed to determine whether a peer is trustworthy. With the first method [3-5], peers only maintain their direct interaction experience with other peers. Before transaction, a peer polls or searches the network for the reputation of the …