Personalized Incremental Credit Computation Model for Distributed System

• Main Authors: Zhi-yang Zheng, 鄭智陽
• Other Authors: 蔡孟峰
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/27133327063397193717

碩士 === 國立中央大學 === 資訊工程學系 === 103 === This paper is designed and applied for the WeOS platform. The WeOS platform is a web app platform based on P2P (Peer-to-Peer) structure network. Users can use the application which is developed by others and they can upload their application as well. The WeOS takes the AID (Autonomous Identity) mechanism and that makes users can manage their identities autonomously. Users can manage their identities and data on the client side with the support of AID. The users’ identities will not store in a central server. In such environment, how can we supervise users’ activity and identify a trusty user without a central server is a hard issue. We design a metric to evaluate a user’s credit according the user’s actions and reputations on the net. Users can select the trusty information source and service provider with the help of credit. The credit will be one of the most important modules in the WeOS. In the future, users’ reactions between service and service should refer to their credit. We think that comment is a subjective action. Everyone can have their own view and it will be very different each other. The traditional reputation models are usually global. Users are not always persuaded by the global value, so we adopt the personal mechanism. We try to make the credit value can express users’ self-experience. However, the personal mechanism will make the computation and storage requirement increase heavily. It is a big loading for a central server to handle that and it is hard to update the value immediately. We take advantage of P2P’s feature of the WeOS platform to separate the task to several nodes. That will share the loads on server and make the procedure can complete rapidly. With the shorter updating cycle, the credit will reveal the users’ behavior more correctly.