基于全同态加密的决策树构造方法  

Decision tree building based on fully homomorphic encryption

作　　者：周李威 王丽珍[1] 张成君[1] 朱玉全[1] 

机构地区：[1]江苏大学计算机科学与通信工程学院,江苏镇江212013

出　　处：《信息技术》2016年第10期28-31,35,共5页Information Technology

基　　金：国家自然科学基金资助项目(71271117);江苏省六大人才高峰项目(2013-WLW-005);江苏省自然科学基金项目(BK20150531)

摘　　要：决策树分类是从一组无次序、无规则的事例中推理出分类规则,从根到每个叶子节点的一条路径就对应着一条分类规则。隐私保护是决策树构造中的一个重要研究课题,任意一个站点都不愿意透露自己的隐秘的数据给其它站点,目前大多数的隐私保护技术都受制于有限的保护凭证和大量的计算耗费。为此,提出一个基于全同态加密协议的决策树构造方法,该协议对数据进行离散化、加密、同态运算。方法允许两个站点之间安全有效的交互各自的数据,达到隐私保护的目的,并能构造出相应的决策树。实验结果表明算法是有效的。Decision tree is a classification rule which is from a set of no order and no rules cases,and there is a path from root node to each children node,the path is a classification rule. Privacy preserving is an important issue in the field of decision tree building,neither of them wants to disclose their private data to the other party or any third party. Most of the existing techniques for privacy preserving suffer from weak privacy guarantees and have a high computational cost involved. Therefore,this paper proposes a protocol based on fully homomorphic encryption that allows two party exchange their data securely and builds a decision tree efficiently. By conducting homomorphic operation with discretization data can protect data safely. The result shows that this protocol achieves good performance.

关 键 词：隐私保护 全同态加密 决策树 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]