Guided Diffusion Model for Adversarial Purification

TL;DR

This paper introduces GDMP, a diffusion-based purification method that effectively defends against adversarial attacks on image classifiers by removing perturbations through guided denoising, significantly enhancing robustness.

Contribution

The paper proposes a novel diffusion-based purification technique, GDMP, integrating adversarial defense into the diffusion process to improve classifier robustness against attacks.

Findings

GDMP reduces adversarial perturbations effectively.

GDMP improves robust accuracy by 5% on CIFAR10.

GDMP achieves 70.94% robustness on ImageNet.

Abstract

With wider application of deep neural networks (DNNs) in various algorithms and frameworks, security threats have become one of the concerns. Adversarial attacks disturb DNN-based image classifiers, in which attackers can intentionally add imperceptible adversarial perturbations on input images to fool the classifiers. In this paper, we propose a novel purification approach, referred to as guided diffusion model for purification (GDMP), to help protect classifiers from adversarial attacks. The core of our approach is to embed purification into the diffusion denoising process of a Denoised Diffusion Probabilistic Model (DDPM), so that its diffusion process could submerge the adversarial perturbations with gradually added Gaussian noises, and both of these noises can be simultaneously removed following a guided denoising process. On our comprehensive experiments across various datasets, the proposed GDMP is shown to reduce the perturbations raised by adversarial attacks to a shallow range, thereby significantly improving the correctness of classification. GDMP improves the robust accuracy by 5%, obtaining 90.1% under PGD attack on the CIFAR10 dataset. Moreover, GDMP achieves 70.94% robustness on the challenging ImageNet dataset.