Privacy-Preserving Collaborative Filtering Using Fully Homomorphic Encryption

Seiya Jumonji, Kazuya Sakai, Min-Te Sun, Wei-Shinn Ku

Abstract: In this paper, we propose privacy-preserving user-based CF protocols using the BGV fully homomorphic encryption scheme, named BGV-CF and optimized BGV-CF (OBGV-CF), in order to protect privacy of users in recommender systems. The proposed schemes are implemented by C++, and testbeds using the MovieLens dataset demonstrate that the proposed protocols successfully achieve their design goals.