Query-Aware User Privacy Protection for LBS over Query-Feature-based Attacks

Massive amount of queries with location data from mobile sensors, connected vehicles, and IoT devices make the Location-Based Services (LBS) pervasive. LBS service providers can provide valuable services for mobile users ranging from POI (Point-Of-Interest) discovery to local search, etc. However, user personal privacy issues mainly come from the fact that users need to send their coordinates and query interests or features to the service provider which may compromise their privacy. In this paper, we focus on a new user privacy problem from a historical query point of view, and consider query search features as new tools for attackers to possibly locate users in a specific location and/or region. We introduce a new query-feature-based inference attack with an illustration on a real-world data set. We define Indistinguishable Feature-Inferred Location/Grids and Probabilistic k-Effectiveness to provide a strong property with differential privacy mannered guarantee. To achieve the privacy property, we design novel randomized algorithms to countermeasure the privacy attack. Based on entropy and recourse cost related metrics, simulations and analysis are performed to show the effectiveness and efficiency of our approaches.