位置轨迹相关性差分隐私保护技术研究与进展

On the Research and Progress of Location Trajectory Correlation Differential Privacy Preservation Techniques

随着移动互联网、智能终端设备和GPS定位技术的普及应用,产生了大量轨迹数据以应用于基于位置的服务.然而,具有时空相关性的行动轨迹可以体现或泄露单用户的个人信息或多用户之间的某些社会关系信息,进一步导致严重的隐私泄露.这些位置轨迹数据的共享和发布离不开隐私保护技术的支持,而大多数隐私保护模型的安全性是基于一定的背景知识,若背景知识已被攻击者掌握,则无法提供完善的隐私保护.差分隐私技术有效弥补这一缺陷,并广泛应用于位置轨迹数据相关性隐私保护领域中.本文首先介绍差分隐私、全局敏感度、噪声机制以及组合性质等相关概念;其次系统地对基于差分隐私的位置轨迹相关性隐私保护技术进行了研究与分析,进一步地详细阐述了差分隐私模型下单条轨迹内相关性保护、两条不同轨迹之间相关性保护以及多条不同轨迹之间相关性保护在时间、空间以及时空方面的隐私保护方法;最后,在对现有方法对比分析的基础上,提出了位置轨迹相关性隐私保护未来的重点研究发展方向.

With the popular application of mobile Internet,smart terminal devices and GPS positioning technology,a large amount of trajectory data have been generated for the application of location-based services.However,spatio-temporal correlated trajectories can reveal personal information of a single user or certain social relationship information among multiple users,which may lead to serious privacy leakage.These location trajectory data can be shared and distributed without privacy preservation,and the security of most privacy preservation models is based on certain background knowledge,which cannot provide perfect privacy preservation if the background knowledge is already in the hands of attackers.Differential privacy technique is effective in bridging this gap and is widely used in the field of privacy preservation of location trajectory data correlation.This paper firstly introduces the concepts of differential privacy,global sensitivity,noise mechanism and combination properties,then it systematically analyzes the location-trajectory correlation privacy preservation techniques based on differential privacy,and further elaborates the privacy preservation methods in terms of time,space and spatio-temporal for the correlation preservation within a single trajectory,between two different trajectories and between multiple different trajectories under the differential privacy model.Finally,based on the comparative analysis of the existing methods,some future key research development directions of location trajectory correlation privacy preservation are proposed.