Privacy preserving in the publication of large-scale trajectory databases

Abstract

In recent years, preserving individual privacy when publishing trajectory data receives increasing attention. However, the existing trajectory data privacy preserving techniques cannot resolve the anonymous issues of large-scale trajectory databases. In traditional clustering constraint based trajectory privacy preserving algorithms, the anonymous groups lack of diversity and they cannot effectively prevent re-clustering attacks against the characteristics of publishing data. In this thesis, a segment clustering based privacy preserving algorithm is proposed. Firstly, the original database is divided into blocks and each block is treated as a separate database. Then, the trajectories in each block are partitioned into segments based on the minimum description length principle. Lastly, these segments are anonymized with cluster-constraint strategy. Experimental results show that the proposed algorithm can improve the safety and have good performance in data quality and anonymous efficiency.