PureGen: Universal Data Purification for Train-Time Poison Defense via Generative Model Dynamics

Omead Pooladzandi, Sunay Gajanan Bhat, Jeffrey Jiang, Alexander Branch, Gregory Pottie

Keywords: Energy-Based Models, Diffusion, Langevin dynamics, Poisons, robustness, defense, Backdoor

TL;DR: A universal, SoTA set of train-time poison defense pre-processing techniques using specially trained Energy Based and Denoising Diffusion Models.

Abstract: Train-time data poisoning attacks threaten machine learning models by introducing adversarial examples during training, leading to misclassification. Current defense methods often reduce generalization performance, are attack-specific, and impose significant training overhead. To address this, we introduce a set of universal data purification methods using a stochastic transform, $\Psi(x)$, realized via iterative Langevin dynamics of Energy-Based Models (EBMs), Denoising Diffusion Probabilistic Models (DDPMs), or both. These approaches purify poisoned data with minimal impact on classifier generalization. Our specially trained EBMs and DDPMs provide state-of-the-art defense against various attacks (including Narcissus, Bullseye Polytope, Gradient Matching) on CIFAR-10, Tiny-ImageNet, and CINIC-10, without needing attack or classifier-specific information. We discuss performance trade-offs and show that our methods remain highly effective even with poisoned or distributionally shifted generative model training data.

Primary Area: Safety in machine learning

Submission Number: 2577