On the Privacy of Horizontally Partitioned Binary Data-Based Privacy-Preserving Collaborative Filtering

Abstract

Collaborative filtering systems provide recommendations for their users. Privacy is not a primary concern in these systems; however, it is an important element for the true user participation. Privacy-preserving collaborative filtering techniques aim to offer privacy measures without neglecting the recommendation accuracy. In general, these systems rely on the data residing on a central server. Studies show that privacy is not protected as much as believed. On the other hand, many e-companies emerge with the advent of the Internet, and these companies might collaborate to offer better recommendations by sharing their data. Thus, partitioned data-based privacy-persevering collaborative filtering schemes have been proposed. In this study, we explore possible attacks on two-party binary privacy-preserving collaborative filtering schemes and evaluate them with respect to privacy performance.