Cryptographic framework for analyzing the privacy of recommender algorithms

Recommender algorithms are widely used, ranging from traditional Video on Demand to a wide variety of Web 2.0 services. Unfortunately, the related privacy concerns have not received much attention. In this paper, we study the privacy concerns associated with recommender algorithms and present a cryptographic security model to formulate the privacy properties. We propose two privacy-preserving content-based recommender algorithms and prove their properties. Moreover, we show the potential weakness in some existing collaborative filtering algorithms which claim to provide privacy protection