Enhancing the Utility of Privacy-Preserving Techniques in Location-based Services (LBS) Applications

Abstract

The widespread use of GPS-enabled devices has led to the proliferation of spatial data (e.g., locations, trajectories), enabling applications like ride-hailing and contact tracing. However, sharing such data raises significant privacy concerns, as sensitive information – such as personal habits or health conditions – can be inferred from spatial patterns. While Differential Privacy (DP) provides rigorous theoretical guarantees for privacy preservation, its noise-injection mechanisms often degrade data utility, limiting the accuracy of location-based services (LBS). Thus, there is an urgent need for privacy-preserving techniques that maintain data utility.

This thesis addresses this challenge by developing novel frameworks that integrate DP with security-based methods (e.g., Secure Multiparty Computation (SMC) and Homomorphic Encryption (HE)) across three critical applications: (1) spatial crowdsourcing, where we propose𝑘-Switch, which achieves 37% improvement in task assignment success rates compared to the baseline; (2) contact tracing, where we introduce ContactGuard, which accelerates SMC operations using Geo-I-perturbed trajectories, maintaining 98% recall in identifying close contacts; and (3) spatial federation, where we develop FedGroup, which reduces the aggregate Laplace noise by 72% compared to other standard DP baselines.

We demonstrate that our frameworks achieve provable privacy guarantees (satisfying 𝜖-differential privacy or its variants) while significantly improving the utility and efficiency over state-of-the-art methods, verified by extensive experiments. The thesis concludes with open challenges and future directions.

Date of Award	2025
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Lei CHEN (Supervisor)