垂直划分多决策表下基于条件信息熵的隐私保护属性约简  

Privacy preserving attribute reduction based on conditional information entropy over vertically partitioned multi-decision tables

作　　者：叶明全[1,2] 胡学钢[1] 伍长荣[3] 

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]皖南医学院计算机教研室,安徽芜湖241002 [3]安徽师范大学数学计算机学院,安徽芜湖241002

出　　处：《山东大学学报（理学版）》2010年第9期14-19,26,共7页Journal of Shandong University(Natural Science)

基　　金：国家自然科学基金资助项目(60975034);安徽省高校省级自然科学研究资助项目(KJ2010B241)

摘　　要：针对垂直划分多决策表,利用半可信第三方和交换加密体制,设计了一个安全多方计算交集基数协议。利用该协议设计了安全多方计算信息熵和安全多方计算条件信息熵的解决方案,提出了一种基于条件信息熵的隐私保护属性约简算法。该算法基于粗糙集信息观的约简理论实现了分布式环境下全局属性约简的求解,使各参与方在不共享其隐私信息的前提下达到集中式属性约简的效果。分析结果表明该算法是有效可行的。A privacy-preserving set intersection cardinality computation protocol based on a semi-trusted third party and commutative encryption is developed, which can be used to solve privacy-preserving computational problems, such as information entropy computation and conditional information entropy computation. A privacy-preserving attribute reduction algorithm based on conditional information entropy for the vertically partitioned multi-decision tables is proposed. The algorithm can globally compute the valid attribute reduction using the attribute reduction idea based on the information viewpoint of Rough set theory, which can get accurate attribute reduction effect on the premise of no sharing of private information among participators. Analysis results show the proposed algorithm is efficient.

关 键 词：属性约简 隐私保护 安全多方计算 粗糙集 条件信息熵 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]