A Fuzzy-logic-based Reputation System On P2P Network

In P2P (Peer-to-peer) network, users can freely connect to each other directly to send and receive file without through a central server. At the same time, the distributed nature of P2P system has brought some potential safety hazards to the stability and usability of system, such as lots of malicious peers and freerider peers. Some reputation systems are invented to settle such problems. But most systems use precise methods to calculate the reputation value which is unable to be precisely described and confirmed in this way due to its characteristics of dynamics, subjectivity and uncertainness. Shanshan Song proposed the FuzzyTrust system for establishing trust in the e-commerce application. FuzzyTrust is based on a fuzzy-logic approach, which can better handle uncertainty, fuzziness and subjective information. FuzzyTrust just focuses on aggregating the peers' reputation, but complete lacks analyzing the behavior of malicious peer and addressing incentive problem.The paper presents a new P2P reputation system in file-sharing application, called FuzzyRep, leveraging the advantages of FuzzyTrust's ability to aggregate the peers' reputation. In order to provide adequate support to coping with peers' strategically altering behavior, FuzzyRep is based on a set of fuzzy logic rules, and uses a formula with the purpose of punishing the malicious peers, to evaluate the peers' reputation value which not only effectively indicates the peer's long term behavior, but also indicates the peer's quick changes. Safety aspects of FuzzyRep are also discussed. In addition, a novel incentive mechanism based on the concept of "birds of a feather flock together" is proposed, in which four parameters in computing peers' contribution value, namely, Shares resources, Downloaded resources, Uploaded resources and Award-penalty factor are introduced. Then implementation the incentive mechanism is described: A peer could organize its neighbor by analyzing other peers' reputation values and contribution values. A peer that wants to maximum its benefit should directly connect to those peers with higher reputation values and contribution values. Thus the peer with high score becomes popular. In this way, peers with similar score will gather together. Then the high score peers will gather together to form a high reliable network area. On the other hand, malicious peers and ffeeriders will be moved to another network area which is far way from the good peers. If the distance between good peers and malicious peers is greater than the message's TTL, the malicious will never response the good peers' requests, thereby, the number of inauthentic files on the network will be decreased. Finally a solution to restrict a peer's uninhibited use of the resources is presented where a peer should meet an interaction threshold to grain access to the resources. As a result, a peer has to increase its reputation value and contribution value to get more services.