Open Access

ARTICLE

A Mix Location Privacy Preservation Method Based on Differential Privacy with Clustering

Fang Liu^*, Xianghui Meng, Jiachen Li, Sibo Guo

School of Information Science and Engineering, Shenyang Ligong University, Shenyang, 110159, China

* Corresponding Author: Fang Liu. Email:

(This article belongs to the Special Issue: Differential Privacy: Techniques, Challenges, and Applications)

Computers, Materials & Continua 2026, 86(2), 1-21. https://doi.org/10.32604/cmc.2025.069243

Received 18 June 2025; Accepted 03 September 2025; Issue published 09 December 2025

Abstract

With the popularization of smart devices, Location-Based Services (LBS) greatly facilitates users’ life, but at the same time brings the risk of users’ location privacy leakage. Existing location privacy protection methods are deficient, failing to reasonably allocate the privacy budget for non-outlier location points and ignoring the critical location information that may be contained in the outlier points, leading to decreased data availability and privacy exposure problems. To address these problems, this paper proposes a Mix Location Privacy Preservation Method Based on Differential Privacy with Clustering (MLDP). The method first utilizes the DBSCAN clustering algorithm to classify location points into non-outliers and outliers. For non-outliers, the scoring function is designed by combining geographic information and semantic information, and the privacy budget is allocated according to the heat intensity of the hotspot area; for outliers, the scoring function is constructed to allocate the privacy budget based on their correlation with the hotspot area. By comprehensively considering the geographic information, semantic information, and correlation with hotspot areas of the location points, a reasonable privacy budget is assigned to each location point, and finally noise is added through the Laplace mechanism to realize privacy protection. Experimental results on two real trajectory datasets, Geolife and T-Drive, show that the MLDP approach significantly improves data availability while effectively protecting location privacy. Compared with the comparison methods, the maximum available data ratio of MLDP is 1. Moreover, compared with the RandomNoise method, its execution time is 0.056–0.061 s longer, and the logRE is 0.12951–0.62194 lower; compared with KemeansDP, QTK-DP, DPK-F, IDP-SC, and DPK-Means-up methods, it saves 0.114–0.296 s in execution time, and the logRE is 0.01112–0.38283 lower.

---

Keywords

---