摘要： 针对数据挖掘中私有信息的保护问题，提出一种多维敏感k-匿名隐私保护模型。将敏感属性泄露问题分为一般泄露、相似泄露、多维独立泄露、交叉泄露和多维混合数据泄露，在k-匿名的基础上，以聚类特性对多维敏感属性进行相似性标记，寻找匿名记录，计算剩余记录与已分组记录的相似性，泛化并发布满足匿名模型的数据集。实验结果表明，该模型适用于多维敏感数据，能防止隐私泄露，数据可用性较好。

关键词: k-匿名, 隐私保护, 多维敏感属性, 属性泄露, 聚类, 相似性

Abstract: This paper focuses on the protection problem of private information in data mining, and proposes multi-dimensional sensitive k-anonymity privacy protection model. Sensitive attribute leakage problems are divided into general leakage similar leakage, multi-dimensional independent leakage, cross leakage and multi-dimensional mixed data leakage. On the basis of k-anonymity, the similarities of multi-dimensional sensitive attribute are marked by clustering features. The model searches for anonymous records, computes the similarities between the remained records and grouped records, and generalizes the dataset satisfied to the anonymous model. The dataset is released. Experimental results show that the model is appropriate for the multi-dimensional sensitive data, and can prevent privacy leaking and have good data availability.

Key words: k-anonymity, privacy protection, multi-dimensional sensitive attribute, attribute leakage, clustering, similarity

中图分类号: 

• TP301.6