Research On Dynamic Trust Management And Evaluation Model In Distributed Environment

Distributed networks have the characteristics of openness, dynamic and voluntary resource sharing. Because there are a large number of fraud behaviors or unreliable services among the massive resources and services, users have to face the problem of how to select an efficient, secure and satisfied resource or service. Currently, one of the effective solutions is to build a trust evaluation system, which can collect and analyze the historical behaviors of entities, and then predict their likely behaviors in the future transactions. Through the trust evaluation system, service requesters select entities, which have higher trust values, to trade. Therefore, it can cut down the loss caused by misunderstanding, reduce the risk of transaction fail, and improve the rate of successful transactions. The research contents are following:1. In the large-scale distributed environment, most of the trust evaluation models have the problems of large overhead and slow rate of trust convergence. Learned from the centralized sales model of similar products and the management functions of organization, we divided the entities in the distributed networks into different trust domains on the basis of application environment and purpose, and propose a hierarchical trust management architecture based on agents and trust domains, which is the system architecture of trust management and metric. Transactions between entities should be completed in their own trust domain as more as possible. This model solves the problems of high management overhead in the distributed environments, inaccurate trust evaluation results in the local trust evaluation models, and large communication overhead and slow trust convergence in the global trust evaluation models. This model also has a large advantage in reducing overhead of distributed management and improving solution efficiency of the system.2. In order to satisfy users’personality preference needs in service selection, combining of service attributes and requesters’personality preference, a fuzzy clustering method based on personality preference is proposed, and used to classify services by the requesters’personality preference. A method to gain the best value of λis proposed and proved in order to improve rationality of the classification. Through these methods, users’service selection happens in the service set that satisfies the users’personality preference, and the success rate and efficiency of service selection improve well. 3. Cloud computing encapsulates and converts computing resources, storage resources and software resources to services, which form a large-scale shared virtual resource pool. And cloud computing has the characteristics of dynamic, uncertainty, distribution and openness. How to select a service which is trusted and satisfy the personality preference from the mass services that have similar or same functions but different QoS. Based on trust and personality preference, a service selection model is proposed in the cloud computing environment. This model is based on agent platform and hierarchical trust management architecture of trust domain, and uses the fuzzy clustering technology based on personality preference. A service selection algorithm is proposed in order to select the closest classification for the requester’s preference. Introducing a trust evaluation mechanism, combining with direct trust and domain recommended trust, a service resource is selected among the requester’s classification, which is secure and trusted, and can satisfy the requester’s personality preference. When the transaction is completed, this model will evaluate the service satisfaction and update the trust degree. The simulation shows that this model can improve the service requesters’satisfaction and has certain resilience to fraud entities.4. The QoS is affected by a variety of factors. After the end of transactions, different users give different evaluated results based on different preferences. Thus users have ambiguity on the evaluated results of the service. The complexity and uncertainty of trust relationships can not be detailed characterized because those trust evaluation standards are not uniform. In order to solve the above problems in trust evaluation systems, a trust evaluation model based on multiple service attributes is proposed. Learning from the transaction procedure in human society, this model extracts multiple service attributes in the distributed environments which can affect the trust evaluation to build an attributes set. The service requester evaluates the service provider’s multiple service attributes and combines with its trust value to determine whether to transact. When the transaction is finished, according to the difference between the actual service quality and the claimed service quality, the service requester will evaluate the trust degree of this service and then update the trust value of the service provider. The simulation shows that the model can accurately evaluate the trust relationships of entities, and effectively curb the attacks of all types of malicious entities. Even there is a higher proportion of the malicious entities in the system, this model still keeps a higher successful transaction rate.5. According to the subjectivity and complexity of the trust relationships, a trust evaluation model based on the extended subjective logic is proposed in order to solve the problem that the evaluation granularity is too rough and the evaluation results are inaccurate in J(?)sang’s models. In order to improve the accuracy and precision of the expression in trust models, this model uses positive trust value [0,1] and negative trust value [0,1] to replace the numbers of positive events and negative events in J(?)sang’s models. Time-attenuation coefficient and transaction value coefficient are introduced to adjust trust evaluation value and improve the rationality of trust value calculation. In order to resist the fraud behaviors of malicious entities, risk factor value is also introduced to calculate the final trust value. The simulation results show that this model can effectively resist the attacks of malicious entities and keep a high system transaction success rate.