Differentially private (DP) machine learning is considered the gold-standard solution for training a model from sensitive data while still preserving privacy. However, a major barrier to achieving this ideal is its sub-optimal privacy-accuracy trade-off, which is particularly visible in DP representation learning. Specifically, it has been shown that under modest privacy budgets, most models learn representations that are not significantly better than hand-crafted features. In this work, we show that effective DP representation learning can be done via image captioning and scaling up to internet-scale multimodal datasets. Through a series of engineering tricks, we successfully train a DP image captioner (DP-Cap) on a 233M subset of LAION-2B from scratch using a reasonable amount of computation, and obtaining unprecedented high-quality image features that can be used in a variety of downstream vision and vision-language tasks. For example, under a privacy budget of ε=8 for the LAION dataset, a linear classifier trained on top of learned DP-Cap features attains 65.8% accuracy on ImageNet-1K, considerably improving the previous SOTA of 56.5%.
@article{arxiv.2403.02506,
title = {Differentially Private Representation Learning via Image Captioning},
author = {Tom Sander and Yaodong Yu and Maziar Sanjabi and Alain Durmus and Yi Ma and Kamalika Chaudhuri and Chuan Guo},
journal= {arXiv preprint arXiv:2403.02506},
year = {2024}
}