1(Department of Computer,Shaoxing University,Shaoxing 312000,China)2(School of Information Management,Nanjing University,Nanjing 210023,China)3(School of Information Engineering,Huzhou University,Huzhou 313000,China)4(Zhejiang Institute of Mechanical and Electrical Engineering,Hangzhou 310053,China)
Abstract：Recommender servicescan guide people to obtain target data from a tremendous amount of resources,and has become an important part of people’ daily life.However,the privacy problem is becoming an important obstacle to the development and application of recommender.Using book recommender as an example,this paper designs and implements a user privacy-preserving book recommender system,whose basic idea is to construct a group of fake profiles carefully for a user profile on a trusted client,to confuse users’ sensitive topics,and thus improve the security of user privacy on the untrusted server.First,a client-based privacy-preserving book recommender framework is presented,which requires no change to the existing book recommender algorithm on the server,and no compromise to the book recommender accuracy.Then,a user privacy protection model is defined,which formally describes the constraints that the fake profiles constructed on the client should meet.Finally,with the help of the book classification catalogue,the implementation of user privacy model is presented.The effectiveness of the system is demonstrated by both theoretical analysis and experimental evaluation.
2020, Vol. 41 
