Extracting Training Data from Diffusion Models
This exposes privacy vulnerabilities in widely used AI models, posing risks for data subjects and requiring new privacy-preserving methods.
The study demonstrated that diffusion models memorize and can reproduce individual training images, extracting over a thousand examples including personal photos and logos, revealing they are less private than GANs.
Image diffusion models such as DALL-E 2, Imagen, and Stable Diffusion have attracted significant attention due to their ability to generate high-quality synthetic images. In this work, we show that diffusion models memorize individual images from their training data and emit them at generation time. With a generate-and-filter pipeline, we extract over a thousand training examples from state-of-the-art models, ranging from photographs of individual people to trademarked company logos. We also train hundreds of diffusion models in various settings to analyze how different modeling and data decisions affect privacy. Overall, our results show that diffusion models are much less private than prior generative models such as GANs, and that mitigating these vulnerabilities may require new advances in privacy-preserving training.