Differentially Private Algorithms for Empirical Machine Learning

By Benjamin Stoddard, Yan Chen, Ashwin Machanavajjhala. 2014

34 citations recorded by the source corpus.

An important use of private data is to build machine learning classifiers. While there is a burgeoning literature on differentially private classification algorithms, we find that they are not practical in real applications due to two reasons. First, existing differentially private classifiers provide poor accuracy on real world datasets. Second, there is no known differentially private algorithm for empirically evaluating the private classifier on a private test dataset. In this paper, we develop differentially private algorithms that mirror real world empirical machine learning workflows. We consider the private classifier training algorithm as a blackbox. We present private algorithms for selecting features that are input to the classifier. Though adding a preprocessing step takes away some of the privacy budget from the actual classification process (thus potentially making it noisier and less accurate), we show that our novel preprocessing techniques significantly increase classifier accuracy on three real-world datasets. We also present the first private algorithms for empirically constructing receiver operating characteristic (ROC) curves on a private test set.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Detecting Violations of Differential Privacy (2018, background): “(b) iSVT 1 [38] adds no noise to query and threshold.”
Understanding the Sparse Vector Technique for Differential Privacy (2016, methodology): “In this paper, we limit our attention to SVT variants to those sa tisfying ǫ-DP, which are what have been used in the data mining community [1, 13, 17, 18].”
On the Privacy Properties of Variants on the Sparse Vector Technique (2015, methodology): “Stoddard et al [9] observed that the privacy guarantee does not depend on ǫ 2 and propose the Private Threshold Testing algorithm that is identical to GPTT with ǫ 1 = ǫ and ǫ 2 = ∞ .”
Pufferfish Privacy Mechanisms for Correlated Data (2016, background): “For the past several years, diﬀerential privacy [8] has emerged as the gold standard in data privacy, and there is a large body of work on diﬀerentially private algorithm design that apply to a growing number of queries [16, 22, 4, 31, 23, 32, 17, 5, 34].”