Experiments and Analyses of Anonymization Mechanisms for Trajectory Data Publishing

Area of Science:

• Computer Science
• Data Privacy
• Geographic Information Systems

Background:

• The proliferation of location-aware devices generates vast amounts of trajectory data, raising significant privacy concerns.
• Existing research presents conflicting findings regarding the re-identifiability of individuals from trajectory data.
• Anonymization techniques for trajectory data often involve a trade-off between data privacy and utility.

Purpose of the Study:

• To systematically evaluate the re-identifiability of individuals from anonymized trajectory data.
• To assess the utility of anonymized trajectory data for practical applications.
• To resolve the ongoing debate on trajectory data privacy and utility.

Main Methods:

• Conducted a systematic experimental study using three real-life trajectory datasets.
• Applied five existing anonymization mechanisms: identifier anonymization, grid-based anonymization, dummy trajectories, k-anonymity, and ε-differential privacy.
• Evaluated data utility through two practical applications: travel time estimation and window range queries.

Main Results:

• Demonstrated that anonymization significantly affects the privacy preservation of trajectory data against re-identification.
• Quantified the utility of anonymized trajectories for specific applications, revealing varying impacts across different mechanisms.
• Provided empirical evidence that effectively closes the debate between conflicting prior studies on trajectory re-identification.

Conclusions:

• The study clarifies the true privacy risks associated with trajectory data and the effectiveness of current anonymization methods.
• Findings highlight the critical need to balance privacy preservation with data utility in trajectory data analysis.
• This research offers a comprehensive framework for understanding and managing privacy in large-scale trajectory datasets.