Research On Trust Evaluation Optimization And Prediction Based On Uncertainty Theory In Cloud Environment

In recent years,cloud computing has become one of the most promising computing models to provide platform,software and infrastructure services on demand through internet technology.In view of its characteristics of virtualization,high scalability,low cost,pay as you go,free from time and space constraints,diversified cloud services attract a large number of small and medium-sized enterprises or individuals to establish business systems or personal applications.Users need to access the virtual service resources stored in the clouds.Traditional information security technology based on "hard security" is difficult to cover its security boundary.Therefore,trust,as a "soft security" technology,is introduced into cloud services security domains to supplement and improve the security management challenges of cloud environment.Due to the difference of personal requirements and the dynamic nature of cloud environment,the trust evaluation,selection and recommendation of cloud services is still a challenge problem.For example,the problems of "cold start" of reputation systems,context-dependence of cloud service performance,the single source of trust evaluation information,the uncertainty of trust,the inhibition of malicious information of nodes,the identification of similar recommenders need to be further studied.Due to the fuzziness and uncertainty of trust,and the dynamic distribution of cloud services,uncertain theory is adopted to study trust evaluation optimization and prediction of cloud services.The main research work are as follows:(1)To better address the context-dependence and subjectivity of trust,trust is decomposed into multiple trustworthiness facets of different importance degree defined by trustors.Meanwhile,owing to the uncertainty,fuzziness of trust and dynamic cloud environment,probabilistic linguistic term set(PLTS)is utilized to measure trust.Furthermore,dynamic reliability and similarity of the recommender's evaluation opinions are taken as the weight information of recommenders,and a flexible and extensible structured trust evaluation framework is established.(2)To better address the dynamic nature of cloud service and the inhibition ofmalicious recommendation information,an extensible trust evaluation model is established from the objective and subjective perspectives.The cloud model theory is utilized to measure the criteria performance from three aspects: criteria average value,variation range and variation frequency.Then the objective criteria evaluation cloud is utilized to identify the similar recommenders,and objective evaluation information of the recommender is used as the benchmark to filter the malicious or biased subjective evaluation information.Meanwhile,the criteria trust grade cloud is established to measure the trust degree of criteria evaluation cloud.Finally,the data fusion rules are employed to aggregate the trust evaluation information of recommenders to obtain the final comprehensive trust value of the target service.(3)Due to the overload characteristics of cloud services,a large number of cloud services with similar functions emerge as the times require.It is impossible for users to invoke all services.Therefore,the accurate Qo S prediction of cloud service is the premise of trust evaluation and recommendation.A cloud service Qo S prediction model which makes full use of the multi-valued time-series Qo S data of specific services is proposed.The model considers the similarity of the time-series data and the consistency of variation trends comprehensively,and recognizes the recommenders who has similar context with the current user,then employ Pareto dominance theory to filter the weak users,retain the dominant users,and achieve accurate prediction of unknown Qo S,providing comprehensive and accurate data support for trust evaluation and selection of cloud services.(4)For the cold start problem of the reputation system of new registered service providers,the reputation bootstrapping model is put forward in our research,and the fuzzy-DEMATEL method is utilized to identify the critical success factors that affect the reputation of service providers,then the neural network is utilized to establish the mapping relationship between the CSFs and the reputation value of the entity to realize the prediction of the reputation value of the new registered entity.