A Reputation System With Dishonest Feedback Detecting Based On Peer-to-peer Network

Peer-to-peer network disaccustomed the former model of permanent server and client. Every peer in the network is completely equal, not only a server, but also a client. Currently, the research of p2p is focused on resources location based on distributed hash table and transaction security based on trust mechanism between peers. The problem of security which is caused by malicious peer behavior is becoming the bottleneck of commercial use of p2p network. The typic malicious behavior involves malicious transact, spurious feedback, traitors, collusion, misguide peers, whitewashers, etc.Reputation system uses certifications signed by client to prove the former behavior of the server, and to calculate its reputation. The Certification System stores certifications which is used to identify server's behavior in the client peer which lastly transact with the server. When a new transact begin, the client asked the previous client to acquire server's certifications. After the transact, the client signs a new certification and adds it to the server's certifications, then stores them. This system could record malicious transact the server conduct efficiently, but is easily attacked by collusion. NICE System stores the certifications in server peer. With the certification, server could easily find the client who trust him, the client which once been a server of other peer could also find peers who trust him. Using this transitive relation, a trust graph could be formed between two strange peers. NICE is performed very well in avoid collusion, but certification which feedback the malicious behavior of server may be throw away by server peer.Our system is a improved reputation system with spurious feedback detect which integrates the advantages of Statement System and NICE System, and adds two kind of new certifications as well as three notions: peer transaction reliability, peer feedback reliability, negative feedback reliability. The improved system contains four basic certifications: positive feedback peer received, negative feedback peer received, positive feedback peer send, negative feedback peer received. Certification of positive feedback peer received is stored in server peer which is used to form the trust graph, so our system obtains the ability of avoid collusion as in NICE. Certification of negative feedback peer received is stored in client peer with which malicious transact of server could be recorded. Peer transaction reliability is the ratio of positive feedback received in all the feedback received. Peer feedback reliability is the ratio of positive feedback peer send in all the feedback peer send. When a feedback is given, the reliability of this feedback is calculated following this rule: If a client with low feedback reliability send a negative feedback to a server with high transaction reliability, the reliability of the negative feedback is low. After the reliability of the feedback is calculated, we use the following approach to update four basic certifications: If a negative feedback gains a high score, the server peer is more likely to conduct malicious transact, server peer should be punished. If the negative feedback gains a low score, the client is more likely to send spurious feedback, client peer should be punished.