Lclean: A Plausible Approach to Individual Trajectory Data Sanitization

TL;DR

Lclean introduces a novel trajectory data sanitization method that protects individual privacy by replacing sensitive points with plausible alternatives, ensuring privacy without compromising data utility.

Contribution

The paper presents a new approach combining location correlation and randomized response to achieve local differential privacy in trajectory data sanitization.

Findings

Protects sensitive trajectory points effectively

Maintains overall data distribution integrity

Ensures local differential privacy

Abstract

In recent years, with the continuous development of significant data industrialization, trajectory data have more and more critical analytical value for urban construction and environmental monitoring. However, the trajectory contains a lot of personal privacy, and rashly publishing trajectory data set will cause serious privacy leakage risk. At present, the privacy protection of trajectory data mainly uses the methods of data anonymity and generalization, without considering the background knowledge of attackers and ignores the risk of adjacent location points may leak sensitive location points. In this paper, based on the above problems, combined with the location correlation of trajectory data, we proposed a plausible replacement method. Firstly, the correlation of trajectory points is proposed to classify the individual trajectories containing sensitive points. Then, according to the relevance of location points and the randomized response mechanism, a reasonable candidate set is selected to replace the sensitive points in the trajectory to satisfy the local differential privacy. Theoretical and experimental results show that the proposed method not only protects the sensitive information of individuals, but also does not affect the overall data distribution.