基于分类敏感属性语义距离的隐私保护模型

Privacy preservation model based on semantic distance of sensitive category attribute

针对传统的k-匿名模型不能有效地解决数据发布中分类敏感属性的相似性攻击问题,提出了新的隐私保护模型——α-similarity k-anonymity模型。由于分类属性值并不是严格意义上有序的,目前缺少一种很好的方法来评价其相似程度。从语义学角度出发,采用分类树的框架来衡量等价类中敏感属性值的语义相似性。实验结果表明,该模型不仅能有效解决身份泄露问题,还能有效解决相似性攻击问题,提供更好的隐私保护效果。

Due to the shortcomings and limitations of these existed privacy preserving approaches in microdata publishing,this paper introduced a novel model named as α-similarity k-anonymity which imposed the additional similarity requirement on categorical sensitive attributes than k-anonymity did.For the categorical attribute,there was short of a method to measure the semantic similarity of sensitive values.This paper adopted the frame of the taxonomy tree to measure the semantic distance among the sensitive values.Experiment results show that the model not only can prevent identity disclosure,but also can prevent similarity attack well.