• •    

DPCS2017+58+基于多敏感属性分级的(α_ij,k, m)-匿名隐私保护方法

王秋月,葛丽娜,耿博,王利娟   

  1. 广西民族大学
  • 收稿日期:2017-07-28 修回日期:2017-09-02 发布日期:2017-09-02
  • 通讯作者: 王秋月

DPCS2017+58+Classification ( ,k, m)-anonymity Privacy Preservation Based on Multi-sensitive Attributes

  • Received:2017-07-28 Revised:2017-09-02 Online:2017-09-02
  • Contact: Qiu-yue WANG

摘要: 针对单敏感属性匿名化存在的局限性和关联攻击的危害问题,本文提出了基于贪心算法的 -匿名模型。首先,该 -匿名模型主要针对多敏感属性信息进行保护,然后,该模型为每个敏感属性的敏感值进行分级设置,有m个敏感属性就有m个分级表,其次,并为每个级别设置一个特定的 ,最后,设计了基于贪心策略的 匿名化算法,采取局部最优方法,实现该模型的思想,提高了对数据的隐私保护程度。并从数据表泛化的信息损失 、执行时间、等价类敏感性距离度量 ,这三个方面对四个模型进行对比。实验结果证明,该模型虽然执行时间稍长,但信息损失量小,对数据的隐私保护程度高,能够抵制关联攻击,保护多敏感属性数据。

关键词: 单敏感属性, 匿名化, 关联攻击, 多敏感属性, (α—ij ,k,m) -匿名模型

Abstract: To solve the problem of the existing limitations of single anonymous sensitive attributes.,An -anonymous model based on greedy algorithm is proposed in this paper. At first, the -anonymous model is mainly for multi-sensitive attribute information, and then, the model for each classification was carried out on the sensitive value of sensitive attribute set, there are m sensitive attributes, and there are m table, and second, set a specific account for each level, and finally, design the -anonymous algorithm based on greedy strategy, adopts the method of local optimum, implement the ideas of the model, to improve the degree of data privacy protection, and compares the four models from the view of degree anonymization cost, difference of ,difference of times, difference of . The experimental results show that although the model has a slightly longer execution time, the information loss is small, and the privacy protection of data is high, which can resist the related attack and protect the data of multi-sensitive property.

Key words: single sensitive attribute, anonymization, associated attack, multi-sensitive attributes, (αij ,k, m)-anonymity models

中图分类号: