一种基于粗糙集属性重要度和密度聚集的匿名化方法  

A Method of Data Anonymization Based on Significant Degree of Quasi Attributes and Density Based Clustering

作　　者：邱桃荣[1] 段文影[1] 段隆振[1] 白小明[1] 

机构地区：[1]南昌大学信息工程学院,江西南昌330031

出　　处：《南昌大学学报（工科版）》2013年第3期297-302,共6页Journal of Nanchang University(Engineering & Technology)

基　　金：国家自然科学基金资助项目(61070139);江西省自然科学基金资助项目(20114BAB201039);江西省科技支撑计划资助项目(20112BBG70087)

摘　　要：基于微聚集技术的k-匿名化MDAV算法没有考虑数据属性的分布情况和数据属性重要性在聚类中的作用,易产生不合理的划分,从而对数据的保护程度与数据可用性之间关系带来影响。针对这个问题本文提出一种基于属性重要度和密度聚类的MDAV改进方法实现对数据集k-匿名化。首先采用基于密度聚类DENCLUE方法对数据表进行聚集成簇,然后对每个簇采用基于粗糙集属性重要度作为加权距离的权值来计算相似样本,实现对数据集的k-划分。与MDAV算法比较测试,所改进的方法改善了发布数据的可用性。Some non-rational clusters was easily generated by Maximum Distance to Average Vector （MDAV） without considering the data distribution and the significance of the quasi-attributes in clustering, which may bring the influence of the trade-off between the utility of anonymized data and privacy protection. In order to deal with the problem, a MDAV algorithm was proposed for modifying of k - anonymity of data. In the proposed method, the clus- ters were generated based on the density method DENCLUE on the given data. And the weighted distance measure, in which the weighted values were obtained by using the quasi attributes significances computed based on Rough set method,was used to implement k - partition in each cluster. Experimental results showed that the k - anonymity based on the proposed MDAV modified method can generate anonymity table improving the utility of anonymized data.

关 键 词：匿名化算法 微聚集 粗糙集 属性重要度 密度聚集 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]