Road Network-Aware and Differentially Private Framework for Location and Location Histogram
Ping Zhao, Jianming Wu, Guanglin Zhang · IEEE Transactions on Intelligent Transportation Systems · 2025
With the development of wireless communication technologies, mobile users can locate themselves and thereby query the untrusted server for Location Based Services (LBS). However, the private information implied by these locations is disclosed to the untrusted server. Limited by computation, communication and storage resources of mobile devices, existing works focused on protecting fine-grained information, namely the snapshot location, or the coarse-grained statistics, i.e., the location histogram, using differential privacy. Nevertheless, preserving the snapshot location cannot prevent the privacy disclosure based on the location histogram, and vice versa. To this end, we propose road network-aware and differentially private framework that can protect both the snapshot location and the location histogram simultaneously. Specifically, we first design Road Network-based Obfuscated Locations Sampling algorithm to sample road networks into discrete locations. Then, we propose Semantic-based Histogram Privacy Protection to elaborately choose discrete locations that satisfy the location histogram differential privacy. Thereafter, we design Road Network-based Differential Privacy Mechanism to perturb these selected discrete locations to protect the user’s snapshot location. Then, we theoretically prove that the proposed framework provides snapshot location ϵ-differential privacy and location histogramc-differential privacy. Finally, the extensive results on four real-world datasets validate the superiority of our work. Specifically, the adversary’s Estimation Error in our work is reduced by 10-12 times compared to the latest work focusing on location histograms, while the adversary’s User Recognition Rate is decreased by 2-5 times compared to the latest work focusing on snapshot locations. Furthermore, our work has excellent performance in terms of Implausible Location Rate, Precision, and Recall.