Differentially private recommender system

Recommender systems is a promising solution to the problem of information overload as they attempt to provide personalized recommendations based on the historical records of users’ activities. Although people’s tastes vary, they do follow patterns so a recommender system can estimate these patterns of taste and discover new and desirable items people didn’t already know. It is an important issue in both academics and industries. The main objective of this chapter is to present the design of a number of recommender systems that are able to provide comprehensive privacy for individuals while minimizing the accuracy loss of recommendations based on differential privacy. In particular, this chapter shows three such applications of differential privacy in providing privacy preserving capabilities for building differentially private recommender systems: (1) differentially private untrustworthy recommender system, (2) differentially private trustworthy recommender system, and (3) private neighborhood-based collaborative filtering (PNCF) method, with an emphasis on the PNCF method.