«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07期

页码:

111-114

栏目:

安全与防范

出版日期:

2017-07-10

文章信息/Info

Title:

 A Decision Tree-based Privacy Preserving Classification Mining Algorithm for Data Streams

文章编号:

1673-629X（2017)07-0111-04

作者:

陈煜; 李玲娟

 南京邮电大学 计算机学院

Author(s):

 CHEN Yu; LI Ling-juan

关键词:

 ; 决策树; 隐私保护; 数据流; 分类

Keywords:

 ; decision tree; privacy preserving; data stream; classification

分类号:

TP311

文献标志码:

A

摘要:

 隐私保护的决策树挖掘方法主要是基于数据扰动的方法和基于安全多方计算的方法.由于数据流高速、连续无限和动态的特性,这些隐私保护方法在数据流挖掘应用上有所不足.针对当前数据流挖掘应用中的隐私泄露问题,提出了一种基于决策树的隐私保护的数据流分类算法-PPFDT.该算法通过采用添加随机噪声的方法对数据加以隐私保护,改进经典的数据流挖掘算法-VFDT,并使用阈值算法找到扰动数据流的最佳分裂属性和最佳分裂点,从而直接在扰动数据流上建立决策树,通过使用该决策树对初始数据流和扰动数据流分类得到较精准的结果.从PPFDT算法的隐私保护程度和在直接扰动的数据流上的分类性能两方面,基于UCI的WaveForm数据集进行了实验验证.实验结果表明,该算法在数据流上快速准确分类的同时,具有一定的隐私保护程度.

Abstract:

 Privacy preserving data mining methods are mainly based on perturbation and randomization approaches and secure multi-party computation approaches.Due to the high-speed data streams with unlimited continuous and dynamic characteristics,these methods are still inadequate.In order to solve privacy leaking problem on current data streams mining application,a privacy preserving fast decision tree mining algorithm for data streams named as PPFDT has been designed and implemented.It adds random noises to protect data privacy and improves the data mining algorithm named VFDT,and uses threshold method to find the best split attribute and the best split point of perturbed data streams,so that a decision tree is directly built on perturbed data streams.Then the decision tree is used to classify original data streams and perturbed data streams for getting accurate results.From the aspects of the privacy protection degree of the PPFDT algorithm and the classification performance on the direct perturbed data stream,the algorithm has been experimentally verified on the Waveform dataset of UCI.The experimental results show that the algorithm can achieve certain degrees of privacy protection,and at the same time,classify data streams fast and accurately.

常用功能

更新日期/Last Update: 2017-08-22