Trust Computing Based On D-S Evidence Theory In Social Networks

Social media,while bringing the convenience and new experience of human communication to people,has indeed triggered an endless stream of moral and legal issues.People's active or passive private data and original digital content in social media are facing huge risks such as data theft,information fraud,privacy snooping and copyright infringement.To protect the privacy of users in social networks,many scholars restrict the communication between users by making complex access control strategies.The most widely used is access control model based on the degree of trust between users.However,with the increase of active users' data in social networks and the increasingly complex social environment,the traditional computing of trust is often difficult to adapt.This topic in the study of the evidence extraction methods affect the user trust social network decision and the basic method of multi-source information fusion for evidence of the trust,this paper proposed a trust calculation method of the improved D-S evidence theory based on social network,this method can better reflect the subjectivity of trust calculation and dynamic,control mechanism,privacy protection method,for social network access content recommendation system to provide strategies based on more accurate application of user trust based on specific innovations are as follows:Predicting the risk privacy messages leakage.Since present trust degree calculation methods cannot predict the risk of leakage in the future,this study predicts the probability of information flow between each user to the black list of the resource owner and takes the distribution as the input that affects the calculation result of the trust evaluation.Individualization of trust evaluation.In present trust calculation methods,the trust decision of all users is computed together.However,the attributes of different users care about when making trust decisions are different.So,it is not reasonable to specify a unified trust calculation model for all users.In this study,we realize the individualization of trust evaluation based on gradient descent algorithm.At the same time,the static attributes of other users are transformed into trust evidence and trust evaluation is based on the combination of trust evidence.Finding a new combination rule.Based on the trust evidence obtained from the first two parts,the final decision of trust degree is made through the combination of evidence.In this study,several existing methods of evidence fusion have been analyzed,and the problem of similarity collision is found.Based on this problem,an evidence fusion method is proposed to reduce the effect that similarity collision impact to the combination result.