文档介绍：第 29 卷第 2 .2
2012 年 2 月 Vol .2012No
Application Research puters Feb
数值型敏感属性的近邻泄露保护方法研究倡
陈伟鹤, 屈洪雪, 邱道龙
(江苏大学计算机科学与通信工程学院, 江苏镇江 212013)
摘要: 针对在发布数值型敏感属性数据时,因同一分组中个体的敏感属性值之间过小的差异而导致攻击者可
以较高的概率以及较小的误差推导出目标个体的敏感信息,从而出现近邻泄露问题,提出了一种有效的防止近
邻泄露的模型:(ε,l)唱。该模型根据不同的敏感属性值区间设置不同的阈值ε(1≤i≤p) 控制敏感属
p anonymity i
性值之间的相似度,并采用有损链接的方法对隐私数据进行保护。实验结果表明,该方法可以明显减少近邻泄
露,提高信息可用性,增强数据发布的安全性。
关键词: 数据发布; 数值型; 有损连接; 可用性; 近邻泄露; (ε,l)唱
p anonymity
中图分类号: 311 文献标志码: 文章编号: 1001唱3695(2012)02唱0650唱05
TP A
: /. .
doi j issn
Research on preservation of proximity privacy of numerical sensitive data
唱, 唱, 唱
CHEN Wei he QU Hong xue QIU Dao long
(School puter Science munication Engineering, Jiangsu University, Zhenjiang Jiangsu 212013, China)
&
Abstract: ’唱
Proximity breach occurs when an. adversary may have high confidence to infer the victim s value fall in a short in
terval in publishing numerical sensitive data In view of such proximity breach in data publishing when sensitive values were
, (εp ,l)唱. 唱
numerical this paper proposed a model called anonymity to prevent proximity breach on the idea of lossy join To con
, εi (1≤i≤p). 唱
trol similarity of sensitive values in different ranges this model set different threshold values The results of ex
periments suggest. that the model is able to reduce proximity breach app