基于指数机制的轨迹差分隐私保护方法

doi:10.11959/j.issn.2096-0271.2022042

摘要/Abstract

摘要：

针对传统轨迹数据保护中忽略位置点携带的语义信息带来的隐私泄露问题，提出一种基于指数机制的轨迹差分隐私保护方法。针对位置空间属性及位置语义特征双重属性信息导致的隐私泄露，根据差分隐私中指数机制的特性，为位置点设计可用的打分函数后随机化输出，对轨迹进行了有效的隐私保护。该方法在保证位置隐私的同时减小数据集规模，并防止语义背景推断攻击，提高数据可用性。在真实轨迹数据集上进行实验，实验结果表明，该方法可以保证隐私强度，有效保护了用户的停留区域位置隐私，同时有效提高了数据可用性。

关键词: 差分隐私, 时空轨迹, 语义位置, 指数机制

Abstract:

A trajectory differential privacy protection method based on exponential mechanism was proposed, aiming at the problem of privacy disclosure caused by ignoring semantic information carried by location points in traditional trajectory data protection.For the privacy disclosure caused by the dual attribute information of geographic features and semantic features of location, an available scoring function for location points was designed according to the characteristics of the index mechanism in differential privacy.And the function randomized the output to protect the trajectory effectively privacy.This scheme could reduce the size of data sets while ensure location privacy, prevent semantic background inference attacks and improve data availability.Experiments were carried out on real trajectory data sets, and the experimental results showed that the proposed method not only effectively protected the privacy of the user's stay area location, but also effectively improved the data availability while ensured the privacy intensity.

Key words: differential privacy, space-time trajectory, semantic location, index mechanism

中图分类号:

TP391

焦荟聪, 刘文菊, 王赜. 基于指数机制的轨迹差分隐私保护方法[J]. 大数据, 2023, 9(1): 141-152.

Huicong JIAO, Wenju LIU, Ze WANG. Trajectory differential privacy protection method based on exponential mechanism[J]. Big Data Research, 2023, 9(1): 141-152.

图/表 9

图1

表1

图2

图3

图4

图5

图6

图7

表2

参考文献 18

[1]	HUO Z , MENG X F , HU H B ,et al. You can walk alone:trajectory privacy-preserving through significant stays protection[C]// Proceedings of International Conference on Database Systems for Advanced Applications. Heidelberg:Springer, 2012: 351-366.
[2]	ZHANG X L , YANG W J . Trajectory anonymous algorithm based on k-means++ against similarity attack[J]. Computer Science and Application, 2020,10(4): 610-618.
[3]	俞望年, 宣占祥, 马小明 ,等. 轨迹数据发布中基于敏感语义位置的隐私保护算法[J]. 现代计算机, 2020(27): 3-9.
	YU W N , XUAN Z X , MA X M ,et al. Privacy protection algorithm based on sensitive semantic location in trajectory data publishing[J]. Modern Computer, 2020(27): 3-9.
[4]	俞庆英, 王燕飞, 叶梓彤 ,等. 基于优化局部抑制的轨迹数据发布隐私保护算法[J]. 计算机工程, 2020,46(8): 112-118.
	YU Q Y , WANG Y F , YE Z T ,et al. Privacy protection algorithm based on optimized local suppression for trajectory data publication[J]. Computer Engineering, 2020,46(8): 112-118.
[5]	王辉, 朱国宇, 申自浩 ,等. 基于用户偏好和位置分布的假位置生成方法[J]. 计算机科学, 2021,48(7): 164-171.
	WANG H , ZHU G Y , SHEN Z H ,et al. Dummy location generation method based on user preference and location distribution[J]. Computer Science, 2021,48(7): 164-171.
[6]	DWORK C . Calibrating noise to sensitivity in private data analysis[J]. Lecture Notes in Computer Science, 2012,3876(8): 265-284.
[7]	HAN Q , XIONG Z , ZHANG K . Research on trajeetory data releasing method via differential privacy based on spetial partition[J]. Security and Communication Networks, 2018: 1-14.
[8]	ZHAO X , PI D , CHEN J . Novel trajectory privacy-preserving method based on prefix tree using differential privacy[J]. Knowledge-based Systems, 2020,198:105940.
[9]	霍峥, 孟小峰 . 一种满足差分隐私的轨迹数据发布方法[J]. 计算机学报, 2018,41(2): 400-412.
	HUO Z , MENG X F . A trajectory data publication method under differential privacy[J]. Chinese Journal of Computers, 2018,41(2): 400-412.
[10]	刘向宇, 刘竹丰, 夏秀峰 ,等. 一种保持轨迹数据高可用性的隐式位置访问隐私保护技术[J]. 沈阳航空航天大学学报, 2019,36(2): 66-75.
	LIU X Y , LIU Z F , XIA X F ,et al. An algorithm of protecting hidden location visit privacy with high trajectory utility[J]. Journal of Shenyang Aerospace University, 2019,36(2): 66-75.
[11]	彭长根, 丁红发, 朱义杰 ,等. 隐私保护的信息熵模型及其度量方法[J]. 软件学报, 2016,27(8): 1891-1903.
	PENG C G , DING H F , ZHU Y J.et al . Information entropy models and privacy protection[J]. Joumal of Software. 2016,27(8): 1891-1903.
[12]	VISVALINGAM M , WHYATT J D . Line generalisation by repeated elimination of points[J]. The Cartographic Journal, 1993,30(1): 46-51.
[13]	MCSHERRY F , TALWAR K . Mechanism design via differential privacy[C]// Proceedings of the 48th Annual IEEE Symposium on Foundations of Computer Science. Piscataway:IEEE Press, 2007: 94-103.
[14]	MYERS C S , RABINER L R . A comparative study of several dynamic time warping algorithms for speech recognition[J]. Bell Labs Technical Journal, 2013,60(7): 1389-1409.
[15]	马明杰, 杜跃进, 李凤华 ,等. 基于语义的位置服务隐私保护综述[J]. 网络与信息安全学报, 2016,2(12): 1-11.
	MA M J , DU Y J , LI F H ,et al. Review of semantic-based privacy-preserving approaches in LBS[J]. Chinese Journal of Network and Information Security, 2016,2(12): 1-11.
[16]	JIANG K F , SHAO D X , BRESSAN S ,et al. Publishing trajectories with differential privacy guarantees[C]// Proceedings of the 25th International Conference on Scientific and Statistical Database Management.[S.l.:s.n.], 2013: 1-12.
[17]	兰微, 林英, 包聆言 ,等. 融入兴趣区域的差分隐私轨迹数据保护方法[J]. 计算机科学与探索, 2020,14(1): 59-72.
	LAN W , LIN Y , BAO L Y ,et al. Trajectory-differential privacyprotection method with interest region[J]. Journal of Frontiers of Computer Science and Technology, 2020,14(1): 59-72.
[18]	LEE J G , HAN J W , WHANG K Y . Trajectory clustering:a partition-and-group framework[C]// Proceedings of 2007 ACM SIGMOD international conference on Management of data.[S.l.:s.n.], 2007: 593-604.

时间	语义
2021-01-02 08:10	XX小区
2021-01-02 08:20	便利店
2021-01-02 09:10	XX医院
2021-01-02 17:10	饭店
2021-01-02 18:30	XX小区
…	…

隐私预算ε	500~ 1 000 m	1 000~ 1 500 m	1 500~ 2 000 m
0.05	0.025011	0.029588	0.015624
0.5	0.025010	0.027752	0.015473
1	0.018261	0.021688	0.012797
5	0.018109	0.022347	0.013085
100	0.016554	0.019989	0.012183
1 000	0.015691	0.019708	0.012155