«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

27

期数:

2017年02期

页码:

81-85

栏目:

安全与防范

出版日期:

2017-02-10

文章信息/Info

Title:

 Investigation on Distributed K-means Clustering Algorithm of Homomorphic Encryption

文章编号:

1673-629X（2017)02-0081-05

作者:

 姚禹丞; 宋玲; 鄂驰

 广西大学计算机与电子信息学院

Author(s):

 YAO Yu-cheng; SONG Ling; E Chi

关键词:

 ; 分布式; K均值聚类; 同态加密; 隐私保护

Keywords:

 ; distributed; K-means; homomorphic encryption; privacy protection

分类号:

TP301.6

文献标志码:

A

摘要:

 针对分布式环境下多方联合执行K均值聚类挖掘任务过程中存在的安全性问题,如潜在的合谋攻击和窃听攻击导致隐私泄露和敏感知识被发现,提出了一种隐私保护算法(PPDK).在数据对象水平分布的情况下,该算法利用同态加密的思想,设计了一种新的加密机制.通过改进加密密钥的生成方式,使得参与计算的各方持有不同的密钥,对于产生的密文,其他参与方无法解密,并且在计算过程中所有的加密解密操作均由各参与方独立完成,因此可以限制半诚实的参与方试图窃听其他参与方的私有信息,以及与中心站点合谋揭露隐私的可能性.通过理论分析和实验结果表明,在有效的时间内,PPDK算法可以在确保分布式K均值聚类挖掘任务得到正确结果的前提下,很好地保护数据的隐私性.

Abstract:

 Aiming at security problems in the process of multi parties performing K-mean clustering mining task under distributed environment,such as potential collusion attacks and eavesdropping attacks leading to privacy disclosure and sensitive knowledge to be found,a privacy protection algorithm,PPDK,is proposed.In the case of the horizontal distribution of the data objects,this algorithm has designed a new encryption mechanism based on the idea of the homomorphic encryption.By improving the generation of the encryption key,it makes each parties hold different keys.One party can’t decrypt the cipher generated by other parties.And in the process of calculating,all encryption and decryption operations are executed by the participants independently.Therefore,it can limit the possibility of semi honest parties trying to eavesdrop the other parties’ private information and conspire with center site.Theoretical analysis and experimental results show that within the effective time,PPDK algorithm can ensure that the distributed K-means clustering mining tasks get a correct results,and the privacy of the data has a very good protection.

常用功能

更新日期/Last Update: 2017-05-11