DynaEgo: Privacy-Preserving Collaborative Filtering Recommender System Based on Social-Aware Differential Privacy

摘要

Collaborative filtering plays an important role in online recommender systems, which provide personalized services to consumers by collecting and analyzing their rating histories. At the same time, such personalization may unfavorably incur privacy leakage, which has motivated the development of privacy-preserving collaborative filtering (PPCF) mechanisms. Most previous research efforts more or less impair the quality of recommendation. In this paper, we propose a social-aware algorithm called DynaEgo to improve the performance of PPCF. DynaEgo utilizes the principle of differential privacy as well as the social relationships to adaptively modify users' rating histories to prevent exact user information from being leaked. Theoretical analysis is provided to validate our scheme. Experiments on a real data set also show that DynaEgo outperforms existent solutions in terms of both privacy protection and recommendation quality.