基于访问概率的位置隐私保护技术

发布时间：2018-07-05 18:32

本文选题：位置隐私 + LBS　；参考：《华东师范大学》2017年硕士论文

【摘要】：近年来,针对位置隐私泄露这一问题,研究者们提出许多新的隐私保护方案。但现有的位置隐私保护技术中很多未考虑访问概率、地图数据、位置语义等边信息,这些信息很可能被恶意攻击者所利用去挖掘用户的位置隐私。即使有些研究者考虑了边信息,他们也将边信息看作是普适的,而不考虑用户的个性化特征。针对LBS中位置隐私保护问题,本文提出了两种新的位置隐私保护方法,分别从全局访问概率和个性化访问概率出发选择位置,从而实现位置k匿名。本文主要工作包括以下几个方面:·基于全局访问概率的位置隐私保护方法该方法首先考虑广大用户的访问特性,从大量的用户访问数据中计算出不同位置的全局访问概率。然后结合用户的隐私需求k,查找满足用户需求的匿名区域,随后计算每个匿名区域的位置信息熵,从中选出信息熵最大的m个区域,并在m个区域中随机选择一个作为用户最终的匿名位置。·基于个性化访问概率的位置隐私保护方法我们提出了一种基于个性化访问概率的位置隐私保护方法,即EPLA方法。对于任一用户,该算法首先根据其在过去已访问的位置计算出这些已访问位置之间距离集合,然后使用核密度估计(Kernel Density Estimation,KDE)计算出用户访问位置之间距离的个性化分布,通过该用户已访问位置之间距离的个性化分布函数计算出每个位置锚点的个性化访问概率。然后我们使用ASS算法选择合适的候选锚点完成该用户位置的个性化匿名。·对EPLA算法优化我们根据核密度估计的计算特征,使用快高斯转换和高斯分布的3σ规则对EPLA进行优化,提出了一种高效的位置隐私保护算法APLA,该算法对EPLA算法中EPVP算法进行优化,很大的改善了 EPLA算法的计算时间开销。
[Abstract]:In recent years, many new privacy protection schemes have been proposed to solve the problem of location privacy disclosure. However, many existing location privacy protection techniques do not consider access probability, map data, location semantics and other edge information, which may be used by malicious attackers to exploit the location privacy of users. Even if some researchers consider edge information, they regard edge information as universal, regardless of the user's personalized features. Aiming at the problem of location privacy protection in LBS, this paper proposes two new location privacy protection methods, which select location from global access probability and personalized access probability, respectively, so as to realize location k anonymity. The main work of this paper includes the following aspects: the location privacy protection method based on global access probability this method first considers the access characteristics of the majority of users and calculates the global access probability of different locations from a large number of user access data. Then, combining the privacy requirement of the user, we find the anonymous region that meets the user's needs, and then calculate the location information entropy of each anonymous region, and select the m regions with the largest information entropy. A location privacy protection method based on personalized access probability is proposed, which is called EPLA method, which is based on personalized access probability. For any user, the algorithm first calculates the set of distances between these visited locations according to the locations they have visited in the past, and then calculates the personalized distribution of the distance between the user's access locations using Kernel density estimation (KDE). The personalized access probability of each anchor point is calculated by the personalized distribution function of the distance between the visited locations. Then we use the ass algorithm to select the appropriate candidate anchor points to complete the personalized anonymity of the user location. We optimize the Gao Si algorithm according to the computational characteristics of the kernel density estimation, and use the fast Gao Si transformation and the 3 蟽 rule of the Gao Si distribution to optimize the Gao Si. This paper presents an efficient location privacy protection algorithm (APLAA), which optimizes the EPVP algorithm in the EPLA algorithm and greatly improves the computational time cost of the EPLA algorithm.
【学位授予单位】：华东师范大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP309

【参考文献】