面向多敏感属性的匿名隐私保护方法

Method of anonymous privacy preserving for multi-sensitive attributes

在数据发布过程中,如果对发布的敏感属性信息不进行任何保护处理而直接发布,容易遭受攻击导致隐私信息泄露。针对传统的单敏感属性隐私保护方法在多敏感属性中不能得到很好的隐私保护效果,提出了一种基于多敏感属性相关性划分的(m,l)-匿名隐私保护模型。利用信息增益法对多敏感属性的相关性进行计算并划分,降低敏感属性维度;根据(m,l)-diversity原则对敏感属性分组,保证发布的数据能防止偏斜性攻击,并且在一定程度上降低背景知识攻击的风险;采用聚类技术实现该模型,减小该模型产生的附加信息损失和隐匿率,确保发布的数据具有较高的可用性。实验结果表明,基于多敏感属性相关性划分的(p,l)-匿名隐私保护模型具有较小的附加信息损失和隐匿率,保证了发布数据的可用性。

In the data publishing process, if the sensitive attribute information is released without any protection process-ing ,it will be vulnerable to be attacked, which leads to the leakage of privacy information. In this paper, since the tradi-tional single-sensitive attribute privacy protection methods do not perform well in the multi-sensitive attributes scenarios, an anonymous privacy protection model based on multi-sensitive attribute relevance partitioning is proposed. First, the infor-mation gaining method is used to calculate the correlation of multi-sensitive attributes, and the dimension of sensitive attrib-utes is reduced. Then, sensitive attributes are grouped according to the （m,l） -diversity principle to ensure that the pub-lished data can prevent skew attacks, and to a certain extent ,the risk of background knowledge attack is reduced. Finally, this model is implemented by clustering technique to reduce the additional information loss and concealment rate of the mod-el and ensure the high availability of the published data. The experimental results show that the anonymity privacy protec-tion model based on multi-sensitive attribute correlation has smaller additional information loss and concealment rate, which ensures the availability of published data.