Research On Privacy-preserving Approach Based On Agent For Cross System In Personalization Service

Cross-system personalized service requires users to agree with share of theirinformation, passes information among different personalized service systems，and thenmaximized integrates and reuses the users’ information. But, users’ information isscattered in various systems, so users’ privacy might be leaked when collaborationcompute happens in different systems. And, because different personalized systems havedifferent user models, matching user profiles is challenging in cross-system personalizedservices, the heterogeneity and privacy preservation of cross-system personalized servicehave become an important and urgent problem. This paper aims at researching theproblem of privacy in cross-system personalized service.First, the paper analyses and compares the advantages and disadvantages about thecurrent approaches of cross-system protection, from the point of view of safety andeffectiveness, proposing an anonymous protection method based on cluster. The newapproach makes partition of members according to the hierarchical clustering algorithmbased on distance, and uses different generalization strategies for different kinds ofidentifiers. The new approach can be effectively reduced the information loss caused bythe generalization while protecting users’ information.Second, this paper researches the user identification and the security for datainformation transfer in several systems, proposing a cross-system personalizedprivacy-preserving approach based on Agent. The implement of user identificationprocesses through a framework which we called Agent. When a user wants to useinformation that comes from other systems, this paper adopts a topic relevance searchmethod which is based on ontology by seeking all similar users in large amounts ofinformation. To solve the problem of unable to determine users’ profiles similarities, thispaper adopt similarity matching algorithm to evaluate the similarities of users.Furthermore, according to the formula of identification probability, the highest similarityusers are determined, finally, we get the same user from the different systems. To solvethe problem of several personalized service systems can secure and accurate acquires user information, this paper adopts a high security algorithm which we calledasymmetrical encryption.Finally, this paper gives the experimental evaluations of the algorithms, comparesand analyses the experimental results with the existing algorithms presently, and makesthe conclusions and prospects for the further research.