一个保护私有信息的布尔关联规则挖掘算法  

An Algorithm for Privacy-preserving Boolean Association Rule Mining

作　　者：罗永龙[1,2] 黄刘生[1,2] 荆巍巍[1,2] 姚亦飞[1,2] 陈国良[1,2] 

机构地区：[1]中国科学技术大学计算机科学技术系,安徽合肥230027 [2]国家高性能计算中心(合肥),安徽合肥230027

出　　处：《电子学报》2005年第5期900-903,共4页Acta Electronica Sinica

基　　金：国家973项目(No.2003CB317000);安徽省教育厅重点科研项目(No.2003kj049zd);安徽省教学研究项目(No.2005166)

摘　　要：　本文基于随机响应技术,提出了一种在保护隐私的关联规则挖掘中对数据进行伪装的方法;设计了在伪装的数据集上进行挖掘的算法;分析了算法的效率.实验结果表明,该算法在伪装的数据集上挖掘出的规则与原始规则相比,相对误差不超过2%,并给出了使得相对误差最小时相关参数的取值.<Abstrcat>In distributed systems,some traditional association rules mining algorithms have been developed with all original data being gathered into a centralized site.However,these algorithms are not fit for the situation where no user is willing to disclose his information.In the privacy preserving association rule mining problems,there are several participants engaged in the computation and the algorithms are run on the union of their databases.Currently,the secure union algorithm can be used to protect each user's privacy if all the user's databases have the same structure.However,in secure union algorithm,each participant should encrypt all the participants' data.So,if there are many participants engaged in the cooperative computation,this method is inefficient.Thus,in this paper,we introduce a data disguised method for privacy preserving association rule mining based on the randomized response techniques,present the mining algorithm on the disguised item set and analyze the complexity of this algorithm.The experiments show that the rule that this algorithm gets has fewer relative error which is less than 2% compared with the original rules.We also give some values of the parameters which make the relative error is the lowest.

关 键 词：数据挖掘 关联规则 随机响应 

分 类 号：TP309[自动化与计算机技术—计算机系统结构]