Abstract Machine Learning (ML) models are ubiquitous in real-world applications and are a
constant focus of research. Modern ML models have become more complex, deeper, and …

There is a growing concern that learned conditional generative models may output samples
that are substantially similar to some copyrighted data $ C $ that was in their training set. We …

Most current approaches for protecting privacy in machine learning (ML) assume that
models exist in a vacuum. Yet, in reality, these models are part of larger systems that include …

Recent developments in deep learning have led to great success in various natural
language processing (NLP) tasks. However, these applications may involve data that …

We study (differentially) private federated learning (FL) of language models. The language
models in cross-device FL are relatively small, which can be trained with meaningful formal …

Artificial intelligence (AI) has seen a tremendous surge in capabilities thanks to the use of
foundation models trained on internet-scale data. On the flip side, the uncurated nature of …

Recently, more and more pre-trained language models are released as a cloud service. It
allows users who lack computing resources to perform inference with a powerful model by …

Pre-training large transformer models with in-domain data improves domain adaptation and
helps gain performance on the domain-specific downstream tasks. However, sharing …

The emerging success of large language models (LLMs) heavily relies on collecting
abundant training data from external (untrusted) sources. Despite substantial efforts devoted …

A well-known algorithm in privacy-preserving ML is differentially private stochastic gradient
descent (DP-SGD). While this algorithm has been evaluated on text and image data, it has …