Home | Publications | WTL+26

TraBiMap: A Transition-Centric Probabilistic Binary Map of Trajectories for Privacy-Aware Analysis

MCML Authors

Matthias Schubert

Prof. Dr.

Core PI

Spatial Artificial Intelligence

Abstract

Analyzing collective mobility data unlocks valuable opportunities in domains such as marketing, traffic, urban planning, disaster management, and public health assessment. However, this data is usually stored per individual, resulting in complex mobility pattern extraction and a high risk of privacy threats. To address this, we propose an alternative discrete trajectory storage in grid cell transitions, which (a) decouples user IDs from temporal patterns and (b) obfuscates single users’ travel patterns by not explicitly storing the location-ID relation but instead applying Bloom filters as an approximate set membership data structure. We show that our approximate approach improves performance for analytical queries with sufficient utility for mobility analysis tasks like assessment of crowdedness, co-movement analyses, and travel pattern extraction, while posing privacy gains compared to raw data processing. The practical usefulness of the TraBiMap is evaluated on three trajectory datasets Character, Roma, and Tdrive, along the dimensions privacy protection level, service quality, anonymization success, candidate result set size, and computational performance.

inproceedings WTL+26

ACM SIGSPATIAL 2026

34th ACM SIGSPATIAL International Conference on Advances in Geographic Information Systems. Riverside, CA, USA, Nov 03-06, 2026. To be published. Preprint available.