PRIVACY PRESERVING LOCATION DATA PUBLISHING: A MACHINE LEARNING APPROACH

Abstract

In an era where mobile and online applications continuously record user location data, safeguarding user privacy during data publication has become increasingly critical. Spatiotemporal trajectory datasets, while valuable for research and decision-making, pose significant privacy risks as adversaries may exploit quasi-identifiers or external data sources to re-identify individuals. Traditional privacy-preserving techniques such as k-anonymity and data perturbation, though helpful, often fall short against modern inference attacks. In this study, we propose a robust Machine Learning-based Anonymization (MLA) framework designed to preserve user privacy while maintaining data utility. The proposed system integrates three core components: clustering using a modified K-Means algorithm, dynamic sequence alignment (Heuristic Clustering), and location generalization to anonymize GPS trajectories. Our model processes real-world datasets including T-Drive and Geolife, and demonstrates improved resistance to probabilistic attacks while preserving trajectory utility and ensuring k-anonymity. Comparative analysis between K-Means and heuristic approaches validates the effectiveness of our system in minimizing information loss and enhancing location data security.