A Privacy Preserving Group Recommender Based on Cooperative Perturbation

Recently, group recommendations is increasingly popular among online service providers (SP), which makes the privacy of users' data collected by SPs for recommendation systems a concern. In this scenario, the privacy protection of individual user in a group is of great importance for a secure group recommendation system. In this paper, a cooperative perturbation privacy-preserving scheme for group recommendation is proposed. The proposed scheme allows users in a group to form privacy protection by mutual perturbations. The preference data from the same user group is confused by the proposed method in order to keep privacy from the SP, while the confused data still preserves utility for preference aggregation at the SP side. In addition, during the data transmission from the user group to the SP, the confused data is further transformed via intragroup mutual perturbation to noise-like data with no utility to malicious attackers. On the SP side, an iterative data exaction method is proposed to recover useful information from the perturbed data for running recommendation algorithm. The effectiveness of the proposed scheme is demonstrated by experimental results.