A General Approach to Adding Differential Privacy to Iterative Training Procedures

TL;DR

This paper presents a modular, practical method for integrating differential privacy into iterative machine learning training, addressing challenges of heterogeneous data and simplifying privacy guarantee computations.

Contribution

It introduces a flexible approach that minimizes changes to training algorithms and extends the Moments Accountant for diverse vector types in privacy-preserving ML.

Findings

Provides a modular framework for differentially private training

Extends Moments Accountant to heterogeneous vector sets

Simplifies privacy guarantee calculations for complex models

Abstract

In this work we address the practical challenges of training machine learning models on privacy-sensitive datasets by introducing a modular approach that minimizes changes to training algorithms, provides a variety of configuration strategies for the privacy mechanism, and then isolates and simplifies the critical logic that computes the final privacy guarantees. A key challenge is that training algorithms often require estimating many different quantities (vectors) from the same set of examples --- for example, gradients of different layers in a deep learning architecture, as well as metrics and batch normalization parameters. Each of these may have different properties like dimensionality, magnitude, and tolerance to noise. By extending previous work on the Moments Accountant for the subsampled Gaussian mechanism, we can provide privacy for such heterogeneous sets of vectors, while also structuring the approach to minimize software engineering challenges.