期刊
NEUROCOMPUTING
卷 472, 期 -, 页码 201-211出版社
ELSEVIER
DOI: 10.1016/j.neucom.2021.04.137
关键词
Trajectory data publishing; Personalized differential privacy; Semantic similarity; Privacy level
资金
- National Natural Science Foundation of China [61872416, 62002104, 62071192, 52031009]
- Fundamental Research Funds for the Central Universities of China [2019kfyXJJS017]
- Key Research and Development Program of Hubei Province [2020BAB120]
- Open Research Project of Hubei Key Laboratory of Intelligent Geo-Information Processing [KLIGIP-2018A03]
This paper proposes an optimal personalized trajectory differential privacy mechanism, which achieves personalized privacy protection through establishing probabilistic mobility model and semantic location matching, and demonstrates a better tradeoff between privacy protection and data utility in experiments.
With the development of location-based applications, more and more trajectory data are collected. Trajectory data often contains users' sensitive information, and direct release it may pose a threat to users' privacy. Differential privacy, as a privacy preserving method with solid mathematical foundation, has been widely used in trajectory data publishing. However, current trajectory data publishing methods based on differential privacy cannot fully realize the personalized privacy protection. In this paper, an optimal personalized trajectory differential privacy mechanism is proposed. Firstly, by establishing the probabilistic mobility model of trajectories, we cluster the locations to achieve semantic location matching between different trajectories. Based on the semantic similarity, we identify the templet trajectory, and propose a privacy level allocation method based on stay-points and frequent sub-trajectories. Then, according to the location matching results, we can automatically identify the privacy level of all locations. Combined with the optimal location differential privacy mechanism, we disturb the location points on the user's trajectory before publishing, where different location privacy levels correspond to different privacy budgets. Experiment results on real-world datasets show that our mechanism provides a better tradeoff between privacy protection and data utility compared with traditional differential privacy methods. (c) 2021 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据