Poster
in
Workshop: Challenges in Deployable Generative AI
Differentially Private Generation of High Fidelity Samples From Diffusion Models
Vikash Sehwag · Ashwinee Panda · Ashwini Pokle · Xinyu Tang · Saeed Mahloujifar · Mung Chiang · Zico Kolter · Prateek Mittal
Keywords: [ Diffusion Models ] [ Privacy leakage ] [ memorization ] [ differential privacy ] [ synthetic data ] [ Generative AI ]
Diffusion based generative models achieve unprecedented image quality but are known to leak private information about the training data. Our goal is to provide provable guarantees on privacy leakage of training data while simultaneously enabling generation of high-fidelity samples. Our proposed approach first non-privately trains an ensemble of diffusion models and then aggregates their prediction to provide privacy guarantees for generated samples. We demonstrate the success of our approach on the MNIST and CIFAR-10.