Peer Discovery With Transitive Trust in Distributed System

Abstract

It is a common case that resources belong to different people who are distributed across the world. So, for making good use of resources, people have to cooperate with each other. Cooperation is important not only in the physical world but also in cyberspace. Computers holding different resources also need to cooperate with others. The problems that people cooperate with people and computers cooperate with computers can be abstracted into a high level one – peers cooperate with other peers in a network. In such problems, the first step to initiate a cooperation is loc-ating other peers – you can not cooperate with a peer when you are not even aware of its existence. The task to locate other peers is called “peer discovery”, it is not an easy task, especially in distributed fashion. Peers need to acquire information of other peers from somewhere, if there is not a central party, the only place to acquire such information is other peers. Malicious peers may provide toxic information to other peers. Therefore, unconditionally trust other peers is very dangerous. For security concern, peers need to find a way to decide who is trustworthy and who is not. This thesis aims to establish trust among peers basing on the historical behaviors of other peers. We believe that by using the established trust, a Peer-to-Peer system will be more resilient to Sybil Attack where attackers create a lot of fake identities to deceive honest peers. Our trust system will be implemented andtested in the peer discovery system of Tribler, which is a distributed system helping people share files.