Research On Context-aware Recommender Systems Based On Anonymous Privacy Protection Model

In the era of Internet,e-commerce,social networks,and other network applications have become an integral part of daily life.A variety of applications for different user's personalized information recommendation has been an important way for the user to access relevant information.However,in order to calculate the accurate recommendation results,it is necessary to collect the user context information for different users.The context-aware recommender system is used to make recommendations for the user with the introduction of environmental information.Context-aware recommender system is more accurate than traditional recommendation system,and has the advantages of pervasive computing and personalization,which can further improve the accuracy of recommendation system and user satisfaction.In context-aware recommender systems,a large number of contextual information is collected.The collection and release of the information is the reason why the recommendation system is double-sided.On one hand,more accurate recommend results are obtained by information collection and application,and on the other hand more users' privacy information will be exposed.If the privacy issues of users are not taken into consideration before collection and publication of the data,users' sensitive information can be easily obtained by the attacker directly or indirectly.If more importance is not attached to related issues of user privacy protection,it is likely that user's personal information will be illegal used and even causes enormous economic and spiritual loss.Therefore,the privacy protection of user information has become a top priority of Internet privacy security.Privacy security is a hot issue in the recommender system and one of the hot research topics.The traditional k-anonymous privacy protection algorithm is a kind of secure and effective algorithm for privacy protection,but it cannot very well against some homogeneity or background knowledge attacks.Aimed at the issue of user information privacy protection in recommendation systems,the main contributions of this dissertation are as follows:(1)We analyzed and summarized related researches as well as related technologies of privacy security in the recommendation system.And the most commonly used k-anonymous algorithms are also discussed.(2)Aimed at the disadvantages of the traditional k-anonymous privacy protection algorithm,we proposed two improved algorithms,which can protect the privacy of users by meeting different anonymous protection conditions of user privacy protection.The first algorithm groups sensitive attributes of user information and sets a degree of privacy protection to each group according to user's sensitive attribute value difference in the level of privacy protection to make an effective protection of the diversity of information privacy protection.The second algorithm protects user privacy properties through micro-aggregation algorithm,and has no specific requirements for quasi identifiers and sensitive attributes of users.(3)The experimental results show that the two algorithms can be both used to protect privacy information of users and meanwhile provide users with accurate personalized recommendation.The two algorithms are aimed at different types of data sets for different clustering and privacy preserving processing and generalization.