Differentially private random decision forests using smooth sensitivity
作者：
Highlights：
• We propose an algorithm that produces a differentially-private random decision forest.
• The decision forest substantially outperforms the current state-of-the-art.
• The algorithm only uses queries that are very insensitive to fluctuations in the data.
• The optimal depth for random decision trees with continuous attributes is calculated.
• Sampling without replacement for each tree is shown to improve prediction accuracy.
摘要
•We propose an algorithm that produces a differentially-private random decision forest.•The decision forest substantially outperforms the current state-of-the-art.•The algorithm only uses queries that are very insensitive to fluctuations in the data.•The optimal depth for random decision trees with continuous attributes is calculated.•Sampling without replacement for each tree is shown to improve prediction accuracy.
论文关键词：Privacy,Data mining,Decision tree,Decision forest,Differential privacy,Smooth sensitivity
论文评审过程：Received 29 September 2016, Revised 12 January 2017, Accepted 25 January 2017, Available online 7 February 2017, Version of Record 14 February 2017.
论文官网地址：https://doi.org/10.1016/j.eswa.2017.01.034
