Saved in:
Bibliographic Details
Main Authors: Huang, Chihan, Alsinglawi, Belal, Al-qudah, Islam
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2508.00552
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911158117597184
author Huang, Chihan
Alsinglawi, Belal
Al-qudah, Islam
author_facet Huang, Chihan
Alsinglawi, Belal
Al-qudah, Islam
contents Recent advances in deep neural networks (DNNs) have led to remarkable success across a wide range of tasks. However, their susceptibility to adversarial perturbations remains a critical vulnerability. Existing diffusion-based adversarial purification methods often require intensive iterative denoising, severely limiting their practical deployment. In this paper, we propose Diffusion Bridge Distillation for Purification (DBLP), a novel and efficient diffusion-based framework for adversarial purification. Central to our approach is a new objective, noise bridge distillation, which constructs a principled alignment between the adversarial noise distribution and the clean data distribution within a latent consistency model (LCM). To further enhance semantic fidelity, we introduce adaptive semantic enhancement, which fuses multi-scale pyramid edge maps as conditioning input to guide the purification process. Extensive experiments across multiple datasets demonstrate that DBLP achieves state-of-the-art (SOTA) robust accuracy, superior image quality, and around 0.2s inference time, marking a significant step toward real-time adversarial purification.
format Preprint
id arxiv_https___arxiv_org_abs_2508_00552
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle DBLP: Noise Bridge Consistency Distillation For Efficient And Reliable Adversarial Purification
Huang, Chihan
Alsinglawi, Belal
Al-qudah, Islam
Computer Vision and Pattern Recognition
Recent advances in deep neural networks (DNNs) have led to remarkable success across a wide range of tasks. However, their susceptibility to adversarial perturbations remains a critical vulnerability. Existing diffusion-based adversarial purification methods often require intensive iterative denoising, severely limiting their practical deployment. In this paper, we propose Diffusion Bridge Distillation for Purification (DBLP), a novel and efficient diffusion-based framework for adversarial purification. Central to our approach is a new objective, noise bridge distillation, which constructs a principled alignment between the adversarial noise distribution and the clean data distribution within a latent consistency model (LCM). To further enhance semantic fidelity, we introduce adaptive semantic enhancement, which fuses multi-scale pyramid edge maps as conditioning input to guide the purification process. Extensive experiments across multiple datasets demonstrate that DBLP achieves state-of-the-art (SOTA) robust accuracy, superior image quality, and around 0.2s inference time, marking a significant step toward real-time adversarial purification.
title DBLP: Noise Bridge Consistency Distillation For Efficient And Reliable Adversarial Purification
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2508.00552