Learning Differentially Private Recurrent Language Models

TL;DR

This paper shows that large recurrent language models can be trained with user-level differential privacy guarantees with minimal impact on accuracy, leveraging federated averaging and privacy accounting techniques.

Contribution

It introduces a method to incorporate user-level differential privacy into federated training of large recurrent language models with negligible utility loss.

Findings

Differential privacy can be achieved with minimal accuracy loss on large datasets.

User-level privacy protection is added to federated averaging algorithms.

Private models perform similarly to non-private models on large datasets.

Abstract

We demonstrate that it is possible to train large recurrent language models with user-level differential privacy guarantees with only a negligible cost in predictive accuracy. Our work builds on recent advances in the training of deep networks on user-partitioned data and privacy accounting for stochastic gradient descent. In particular, we add user-level privacy protection to the federated averaging algorithm, which makes "large step" updates from user-level data. Our work demonstrates that given a dataset with a sufficiently large number of users (a requirement easily met by even small internet-scale datasets), achieving differential privacy comes at the cost of increased computation, rather than in decreased utility as in most prior work. We find that our private LSTM language models are quantitatively and qualitatively similar to un-noised models when trained on a large dataset.