Nearly most influential location selection with differentially private user locations in a road network

Abstract

During the past decades, maximum influential location selection (Max-inf) problems have been of intense interest to the spatial database community. The Max-inf problem searches for a location that attracts as many clients as possible, so it is essential to collect the location information of each client for such a query. However, the client location is considered sensitive information, and location privacy has become an emerging issue. To resolve the privacy issue, we present a novel Max-inf problem in a differentially private manner, which is called DP-Max-inf in a road network. Differential privacy is a de-facto standard privacy protection technique that injects controlled noise into statistical query results. In addition, we present the influence region overlapping problem while applying differential privacy to the Max-inf problem using the conventional approach. To remedy this problem, we propose a network Voronoi region-based technique to guarantee query accuracy and a network Voronoi envelope-based pruning heuristic to improve query performance.