Privacy Distillation: Reducing Re-identification Risk of Multimodal Diffusion Models
Abstract
Knowledge distillation in neural networks refers to compressing a large model or dataset into a smaller version of itself. We introduce Privacy Distillation, a framework that allows a text-to-image generative model to teach another model without exposing it to identifiable data. Here, we are interested in the privacy issue faced by a data provider who wishes to share their data via a multimodal generative model. A question that immediately arises is ``How can a data provider ensure that the generative model is not leaking identifiable information about a patient?''. Our solution consists of (1) training a first diffusion model on real data (2) generating a synthetic dataset using this model and filtering it to exclude images with a re-identifiability risk (3) training a second diffusion model on the filtered synthetic data only. We showcase that datasets sampled from models trained with privacy distillation can effectively reduce re-identification risk whilst maintaining downstream performance.
Cite
@article{arxiv.2306.01322,
title = {Privacy Distillation: Reducing Re-identification Risk of Multimodal Diffusion Models},
author = {Virginia Fernandez and Pedro Sanchez and Walter Hugo Lopez Pinaya and Grzegorz Jacenków and Sotirios A. Tsaftaris and Jorge Cardoso},
journal= {arXiv preprint arXiv:2306.01322},
year = {2023}
}