Preserving privacy in collaborative filtering through distributed aggregation of offline profiles

Abstract

In recommender systems, usually, a central server needs to have access to users' profiles in order to generate useful recommendations. Having this access, however, undermines the users' privacy.