Differential privacy-based trajectory community recommendation in social network

Trajectory community recommendation (TCR) is a location-based social network (LBSN) service whereby LBSN server recommends a user a community in which the trajectories have similar movement patterns with the user's trajectory. Due to privacy concerns, the trajectory should be protected. However, the data availability of traditional privacy-preserving schemes is limited, and previous differential privacy (DP) methods cannot achieve high data utility in TCR and rely on a fully trusted third party. In this paper, we propose a DP-based trajectory community recommendation (DPTCR) scheme to perform effective TCR service while protecting trajectory privacy by the user himself. First, DPTCR transits the actual trajectory's locations into noisy feature locations based on private semantic expectation method, which ensures the semantic similarity between noisy locations and the actual locations. Second, DPTCR uses a private geographical distance method to construct a noisy trajectory that has the smallest geographical distance with the actual trajectory. Finally, DPTCR uses a semantic-geographical distance model to cluster a community in which the trajectories have high similarity with the constructed noisy trajectory. Security analysis proves that our DPTCR scheme achieves E-DP, and experimental results show that our scheme achieves high efficiency and data utility. (C) 2019 Elsevier Inc. All rights reserved.