A General Approach to Adding Differential Privacy to Iterative Training Procedures
This addresses privacy-sensitive dataset training for machine learning practitioners, offering an incremental improvement over existing methods.
The paper tackles the challenge of adding differential privacy to iterative machine learning training procedures by introducing a modular approach that minimizes algorithm changes and simplifies privacy guarantee computation, extending the Moments Accountant for heterogeneous vectors like gradients and batch normalization parameters.
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.