Peer-to-peer Network Computing Reputation-based Incentive Model Study

P2P network has been one of the hottest research issues for its attractive advantages in the application areas like file sharing, content distribution, distributed storage, etc. However, the autonomous nature of nodes in P2P network leads to the Free-riding problem and the trust problem in P2P service management, which seriously impact the service availability of P2P environment. The selfishness of free-riders may lead to the lack of service resources. On the other hand, the lack of trust in P2P environment leads to the deterioration of service quality in P2P network. This dissertation concentrates on alleviating the negative impact of node autonomy on the service availability in P2P network, through the design of the trust management mechanism and the incentive mechanism for nodes. The main contributions of this dissertation are as follows:In Peer-to-Peer environments, often require users to utilize other resources that are unknown. Users are unable to identify whether a resource will be honest or malicious. In order for users to reduce the risk of using a malicious resource, we propose a reputation-based trust framework for distributed system applications. In this framework, users are able to appraise a resource by using price and a quantifiable metric of trust that is gathered from its own view and the views of other peers regarding the reputation of the resource. The appraisal process provides the user with a reliable metric that can be used in the process of resource scheduling and selection. The simulation results show that the framework is user friendly by providing long-term high satisfaction, filters out malicious nodes.The anarchic property of P2P system keeps participants uncontrolled. Then, the behavior of free-riding happens. This thesis uses dynamic and incomplete information game theory to explore what results in free-riding. We use the concept of game theory and probability to design an incentive framework to solve free-riding problem, and also propose fairness index to monitor the efficiency of whole system. Finally, we use simulations to testify our framework in static environment. And we choose mechanisms which have no incentive function to be compared with our incentive framework. From the simulation results, our incentive framework can efficiently lower free-riding behavior, and guarantee the fairness of non-free-riders under the same situation.