Research And Application Of Service Reputation Evaluation Based On Cloud Model

Web service as an implementation of service-oriented architecture (Service-Oriented Architecture, SOA) applies to heterogeneous architecture of the Internetquickly. However, significant increase of the number of Web service has brought anew problem, that how to choose the satisfactory Web service for users. Currently,most researches on the problem of service discovery and selection are based onservice reputation using quantization parameters. However the evaluation of servicereputation is inaccurate because of abnormal users and fake QoS declarations, and itis very important to create a model with comprehensive description of servicereputation.Cloud model can give comprehensive description of fuzzy concepts, and reflectcharacteristics of entities from multiple angles. Cloud model expressescharacteristics of qualitative concepts using distribution of cloud droplets andsamples, and reflects true level and stability of entities with cloud parameters Ex, Enand He. Therefore attacks of abnormal users and fake service information can bedetected. Cloud model is introduced into the evaluation of reputation, and theparameter information can be fully retained in the process, and limitations of singlenumerical expression can be avoided. As a result, the influence of abnormalfeedback attacks and fake QoS declarations on reputation evaluation is limited.Specific works are as follows:Firstly, subjective reputation evaluation model based on user feedbacks isestablished. This paper analyzed input samples of cloud model, and proposed afeedback similarity algorithm based on PeerTrust model and user history ratings.With similarities as input cloud droplets, user ratings quality cloud can reflect theauthenticity and stability of user feedbacks. So that feedbacks with low quality willbe eliminated, and users with fluctuations in behavior will be punished. So the resultof reputation evaluation will be more accurate, and it can fight against abnormal userattacks with fake feedback effectively. Secondly, objective reputation evaluation model based on QoS declarationcredibility is established. This paper added up input droplet samples of cloud model,and proposed a similarity algorithm based on QoS declaration and true value. Cloudmodel can reflect authenticity and stability of QoS declaration through servicedeclaration quality cloud. For the services with fake QoS declaration, its reputationwill be punished and reduced, and the comprehensive service reputation will be usedas a basis for service selection.Finally, this paper compares cloud-based reputation evaluation method with theother two methods. According to the result of simulated tests of different scenariosin the simulation environment, the proposed method in fake information resistance,volatile behavior identification and transaction success rate has better performance,which proves the effectiveness of proposed method.