Research On Reputaion-based Trust Mechanism In Open Network Environment

Trust assessment technology is used to dynamically analyze the trust of entities in the network,monitor and prevent malicious or false behavior.Current network systems mostly adopt distributed structure,but in distributed network systems,there is no centralized management authority to rely on,so it is difficult to judge the entity credibility.Although the trust evaluation model currently used has some dynamic applicability,how to deal with false information and malicious attacks in the network and how to measure the credibility is still a problem.In order to reduce the complexity of judging the credibility of entities in the network and finally build a secure and reliable network environment,this dissertation makes an in-depth analysis of a series of key technologies in the evaluation of individual credibility in MAS.The following results have been achieved:1.Aiming at the problem that malicious evaluation may exist in service evaluation data,which leads to low credibility of reputation evaluation results,this dissertation proposes a feedback detection method for malicious evaluation.As malicious evaluation is scattered in evaluation data sets,this dissertation proposes a Q-CUSUM algorithm based on CUUM algorithm,which prepossess the feedback evaluation information to detect malicious feedback.The basic idea of this method is to remove malicious scores from scoring data sets by detecting the changes of feedback evaluation values in time series,so as to ensure the authenticity of scoring data.2.In view of the insufficient use of feedback information in credibility evaluation,this dissertation proposes a feedback information integration method,which divides the content of credibility evaluation feedback into five different dimensions,and then uses the EWM(entropy weight method)method to calculate the weights of the corresponding evaluation of different dimension indicators.By assigning different weights to the evaluation information of different dimensions,the evaluation results reflect the relationship between different evaluation indicators,and combine the utility function to get the integrated results of evaluation feedback.3.Current research lacks specific methods to distinguish the influence of agents who give credit evaluation.Therefore,in order to improve the reliability of reputation evaluation results,the differences of feedback influence among different agents are considered.In this dissertation,a feedback influence evaluation method is proposed.As an auxiliary criterion for credibility evaluation,the interaction records of agents in MAS environment are established,and the interaction links between agents in interactive networks are analyzed.The influence of agents irrelevant to the evaluation content is obtained,which enhances the objectivity of reputation evaluation results.