基于时空特征的差分隐私保护方法  

Differential privacy protection method based on spatiotemporal correlation

作　　者：乔雨 嵇浩 QIAO Yu;JI Hao(Pujiang Institute,Nanjing Tech University,Nanjing 211100,China;Pin Duoduo Information Technology Company Limited,Shanghai 200050,China)

机构地区：[1]南京工业大学浦江学院,江苏南京211100 [2]拼多多信息技术有限公司,上海200050

出　　处：《辽宁工程技术大学学报（自然科学版）》2024年第4期506-512,共7页Journal of Liaoning Technical University (Natural Science)

基　　金：江苏省高等学校基础科学(自然科学)研究项目(21KJB520028);南京工业大学浦江学院自然科学立项项目(njpj2022-1-28)。

摘　　要：为了更好地对轨迹数据中蕴含的个人敏感信息进行保护,采用理论研究与实验证明相结合的方法,在现有的隐私保护框架基础之上进一步考虑时间特性与空间特征之间的关联关系,提出了一种基于时空关联特征的差分隐私保护方法(SC-DP)。根据数据集的局部动态自适应分配时间和空间维度的权重,使得在连续的ε时间窗口以及m×m区域内出现的轨迹子流都能满足ε-差分隐私的要求,并在真实数据集上验证了算法的可用性和实用性,在Geolife数据集上的平均误差降低32.8%,CabSpotting数据集中的性能提升29.2%。In order to better protect the personal sensitive information embedded in trajectory data,an approach builds upon existing privacy protection frameworks and further investigates the correlation between temporal characteristics and spatial features is proposed by combining theoretical research with empirical validation,which named differential privacy protection method based on spatiotemporal correlation features(SC-DP).The method adaptively allocates weights of time and space dimensions according to the local dynamic of trajectories,ensuring the sub-stream appearing in a continuousε-time window and m×m regions that can meet the requirements ofε-differential privacy protection.The algorithm's feasibility and practicality were validated on real datasets,achieving a 32.8%reduction in average error in the Geolife dataset and a 29.2%performance improvement in the CabSpotting dataset.

关 键 词：时空关联特征 轨迹子流 位置服务 差分隐私 数据可用性 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]