K-匿名隐私保护模型中不确定性数据的建模问题研究

建模是不确定性数据管理的基础,K-匿名隐私保护模型中不确定性数据有其特殊性:它是人为泛化后的不确定性数据,泛化后的每个实例还原成泛化前元组的概率是相等的。由于其特殊性,以往针对非人为造成不确定性的数据建模方法已经不能简单地用于描述K-匿名隐私保护模型中不确定性数据。为了描述K-匿名隐私保护模型中不确定性数据,本文提出几种针对它的新建模方法:Kattr模型使用attrib-ute-ors方法来描述K-匿名数据中准标识符属性值的不确定性;Ktuple模型把K-匿名表不确定属性值看成是一个关系值,对关系值使用tuple-ors方法来描述;Kupperlower模型把K-匿名表泛化值范围分开成两个字段:上限和下限;Ktree模型根据K-匿名表是对普通表通过泛化树泛化而形成这一特性逆向拆分成树形结构。由这几种模型及它们之间的组合构成了一个描述K-匿名隐私保护模型中不确定性数据的模型空间。并且,本文讨论了模型空间里各种模型的完备性和封闭性等性质。

Modeling the Uncertain Data in the K-Anonymity Privacy Protection Model

Modeling is the basis for the data management of uncertainty.The specificity in the uncertainty of the data in the k-anonymity privacy protection model is found,namely,its uncertainty is caused by artificial generalization,and the probability that each instance is reduced after generalization to the original tuple is equal.Because of its specificity,the past modeling approaches of uncertainty data are not suitable for the uncertainty data in the k-anonymity privacy protection model simply.In order to describe uncertainty data in the k-anonymity privacy protection model,several new modeling methods are proposed in this paper:the Kattr model uses the attribute-ors ways to describe the uncertainty in the quasi-identifier attribute values of the k-anonymity privacy protection model;the Ktuple model takes the quasi-identifier attribute values as relations and use the tuple-ors ways to describe the relations;the Kupperlower model separates some generalization values to two fields:the upper limit and the lower limit;the Ktree model based on the property that k-anonymous table is the generalization of the ordinary relation with generalization tree splits the quasi-identifier attribute value into a certain tree reversely.A model space which consists of these models is given.The completeness and closure about these models are discussed later.