PPREC: A Privacy Preserving Recommender System with Local Differential Privacy

Abstract

Recommender systems plays an important role today and are widely applied to a wide range of domains. Similar to most machine learning algorithms, recommenda tion systems rely on users’s data to train and to generate recommendations .Often, these data can be private and in case of a leakage, it could seriously cause harm to the users, making it a crucial mission to investigate privacy threats in recom mender systems and implement defensive techniques to address them. In this work we highlight various recommendation systems types and techniques, then we present popular privacy preserving machine learning techniques. Moreover, we revisit the different proposed techniques to preserve privacy in literature. Furthermore, we pro pose PPREC, a privacy preserving recommendation engine using local differential privacy and matrix factorization. Finally, we put the proposed library on numerous tests and discussed the obtained results.