Embedding differential privacy in decision tree algorithm with different depths

Differential privacy（DP） has become one of the most important solutions for privacy protection in recent years. Previous studies have shown that prediction accuracy usually increases as more data mining（DM）logic is considered in the DP implementation. However, although one-step DM computation for de...

Full description

Saved in:
Bibliographic Details
Published inScience China. Information sciences Vol. 60; no. 8; pp. 132 - 146
Main Authors Bai, Xuanyu, Yao, Jianguo, Yuan, Mingxuan, Deng, Ke, Xie, Xike, Guan, Haibing
Format Journal Article
LanguageEnglish
Published Beijing Science China Press 01.08.2017
Springer Nature B.V
Subjects
Accuracy
Algorithms
Complexity
Computer Science
Data mining
Decision trees
DT模型
Embedding
Information Systems and Communication Service
Markov chains
Privacy
Research Paper
Solution space
决策树算法
嵌入
数据挖掘
求解空间
隐私保护
预测精度
马尔可夫链
exponential mechanism
MCMC
decision tree
rential privacy
exhaustive search
Online AccessGet full text

Cover

More Information
Summary:Differential privacy（DP） has become one of the most important solutions for privacy protection in recent years. Previous studies have shown that prediction accuracy usually increases as more data mining（DM）logic is considered in the DP implementation. However, although one-step DM computation for decision tree（DT） model has been investigated, existing research has not studied the scenarios when the DP is embedded in two-step DM computation, three-step DM computation until the whole model DM computation. It is very challenging to embed DP in more than two steps of DM computation since the solution space exponentially increases with the increase of computational complexity. In this work, we propose algorithms by making use of Markov Chain Monte Carlo（MCMC） method, which can efficiently search a computationally infeasible space to embed DP into DT generation algorithm. We compare the performance when embedding DP in DT with different depths, i.e., one-step DM computation（previous work）, two-step, three-step and the whole model. We find that the deep combination of DP and DT does help to increase the prediction accuracy. However, when the privacy budget is very large（e.g., ∈ = 10）, this may overwhelm the complexity of DT model, and the increasing trend is not obvious. We also find that the prediction accuracy decreases with the increase of model complexity.
Bibliography:differential privacy, decision tree, exponential mechanism, exhaustive search, MCMC
11-5847/TP
Differential privacy（DP） has become one of the most important solutions for privacy protection in recent years. Previous studies have shown that prediction accuracy usually increases as more data mining（DM）logic is considered in the DP implementation. However, although one-step DM computation for decision tree（DT） model has been investigated, existing research has not studied the scenarios when the DP is embedded in two-step DM computation, three-step DM computation until the whole model DM computation. It is very challenging to embed DP in more than two steps of DM computation since the solution space exponentially increases with the increase of computational complexity. In this work, we propose algorithms by making use of Markov Chain Monte Carlo（MCMC） method, which can efficiently search a computationally infeasible space to embed DP into DT generation algorithm. We compare the performance when embedding DP in DT with different depths, i.e., one-step DM computation（previous work）, two-step, three-step and the whole model. We find that the deep combination of DP and DT does help to increase the prediction accuracy. However, when the privacy budget is very large（e.g., ∈ = 10）, this may overwhelm the complexity of DT model, and the increasing trend is not obvious. We also find that the prediction accuracy decreases with the increase of model complexity.
ISSN:1674-733X
1869-1919
DOI:10.1007/s11432-016-0442-1