Diffusion Model-Based Effective Privacy-Preserving Computation of LBS User Density Distributions

Location-based services (LBS) have become integral to modern life, supporting diverse applications ranging from navigation and ride-sharing to disaster alerts. To access these services, users must share their location information, which introduces significant privacy risks. Various privacy-preserving methods have been developed to address this concern, with Geo-Indistinguishability (Geo-I) being the most widely adopted. Geo-I protects user privacy by perturbing location data, but this perturbation poses challenges for LBS providers, as it limits their ability to leverage user data for service improvement. To address this issue, this paper proposes a novel method for accurately estimating user density distributions from Geo-I-perturbed location data. The core contribution lies in the use of a diffusion model, an advanced AI-based generative technique known for its success in diverse applications. Experimental results on real-world datasets demonstrate that the proposed approach effectively recovers accurate user density distributions, enabling improved utility of LBS data while preserving user privacy.