电子学报 ›› 2023, Vol. 51 ›› Issue (3): 552-563.DOI: 10.12263/DZXB.20210631

• 学术论文 • 上一篇    下一篇

基于差分隐私的活动模式保护与时空轨迹发布方法

曾卓, 汪成亮, 马飞   

  1. 重庆大学计算机学院,重庆 400044
  • 收稿日期:2021-05-18 修回日期:2022-04-16 出版日期:2023-03-25
    • 通讯作者:
    • 汪成亮
    • 作者简介:
    • 曾卓 男,1991年5月出生,重庆涪陵人.重庆大学博士生,主要研究轨迹隐私保护,动态图神经网络安全等.E-mail: zengz@ cqu.edu.cn
      汪成亮(通讯作者) 男,1975年5月出生,四川资阳人.博士,现为重庆大学计算机学院教授,博士生导师.主要研究领域为复杂系统智能控制,无线网络及RFID研究与应用等.
      马飞 男,1993年4月出生,河南商丘人.重庆大学博士生,主要研究方向为语音信号处理、迁移学习.E-mail: mafei@cqu.edu.cn
    • 基金资助:
    • 国家自然科学基金(61672115);重庆市技术创新与应用发展专项重大主题专项(cstc2020jscx-dxwtBX0055)

Differentially Private Activity Pattern and Spatial-Temporal Trajectory Publication

ZENG Zhuo, WANG Cheng-liang, MA Fei   

  1. Computer School,Chongqing University,Chongqing 400044,China
  • Received:2021-05-18 Revised:2022-04-16 Online:2023-03-25 Published:2023-04-20
    • Corresponding author:
    • WANG Cheng-liang
    • Supported by:
    • National Natural Science Foundation of China(61672115);Chongqing Technology Innovation & Application Development Key Project(cstc2020jscx-dxwtBX0055)

摘要:

为了解决用户轨迹数据发布时的活动模式泄露问题,本文提出了一种基于差分隐私的活动模式保护与时空数据发布方法DPAP-STTP(Differentially Private Activity Pattern and Spatial-Temporal Trajectory Publication),该方法即保护了用户时空数据中活动模式的隐私,又可以保证所发布时空轨迹在服务建议生成上的有效性.在DPAP-STTP中,用户的活动模式表示为个人代表性轨迹的动静态信息,包括代表性轨迹的时空密度分布、时空路径分布、移动模式以及时空跨度.另外,DPAP-STTP通过隐私保护预算与隐私保护阈值对该动静态信息进行调控,然后根据调控后的动静态信息依次划分时空网格、重构轨迹所处时空区间、时空轨迹点随机采样,最终生成满足群体差分隐私的时空轨迹进行发布.本文的实验比较了DPAP-STTP与DP-STAR(Differential Private Synthetic Trajectory Publisher)、BNA(Bounded Noise-Adding)所生成的轨迹在特定时空范围内的有效性,证明DPAP-STTP不但可重构服从群体差分隐私的时空轨迹,而且在时空网格上维持了时空轨迹的有效性.

关键词: 活动模式, 群体差分隐私, 时空轨迹, 动静态信息

Abstract:

In order to solve activity pattern leakage problems while user trajectory data publishing, the paper proposes the DPAP-STTP (Differentially Private Activity Pattern and Spatial-Temporal Trajectory Publication) method to publish spatial-temporal trajectories for achieving required services suggestions in support of users while preserving the privacy of activity patterns. In DPAP-STTP, users' activity patterns are represented as dynamic and static information of personal representative trajectories, including spatial-temporal density distribution, spatial-temporal trip distribution, mobility pattern and spatial-temporal span. Additionally, according to allocated privacy budget and specific privacy-preserving threshold, DPAP-STTP preserves the privacy of dynamic and static information, and uses perturbed information to divide spatial-temporal grids, reconstruct spatial-temporal passing grids, randomly select spatial-temporal point, and finally generate spatial-temporal trajectories with group differential privacy satisfied. The experiment in this paper compares the DPAP-STTP with DP-STAR (Differential Private Synthetic Trajectory Publisher) and BNA (Bounded Noise-Adding) for presenting the utility of DPAP-STTP trajectories. Consequently, the DPAP-STTP method is proved to generate spatial-temporal trajectories which follow the group differential privacy and maintain their utility in some spatial-temporal scopes.

Key words: activity pattern, group differential privacy, spatial-temporal trajectories, dynamic and static information

中图分类号: