DiffDefense: Defending against Adversarial Attacks via Diffusion Models

Hondamunige Prasanna Silva, Lorenzo Seidenari, Alberto Del Bimbo
9/7/2023
cs.LGcs.CRcs.CV

Official code: HondamunigePrasannaSilva/DiffDefence (20 stars, last commit 2024-02-01). 1 more implementation below.

Abstract

This paper presents a novel reconstruction method that leverages Diffusion Models to protect machine learning classifiers against adversarial attacks, all without requiring any modifications to the classifiers themselves. The susceptibility of machine learning models to minor input perturbations renders them vulnerable to adversarial attacks. While diffusion-based methods are typically disregarded for adversarial defense due to their slow reverse process, this paper demonstrates that our proposed method offers robustness against adversarial threats while preserving clean accuracy, speed, and plug-and-play compatibility. Code at: https://github.com/HondamunigePrasannaSilva/DiffDefence.

View on arXivView PDF

Code Implementations(2)

HondamunigePrasannaSilva/DiffDefenceOfficial

Official implementation of the paper DiffDefence: defending against adversarial attacks via diffusion models. ICIAP 2023.

202PythonJul 13, 20232 years ago
adversarial-attacksadversarial-defencediffusion-models
ankile/Adversarial-DiffusionOfficial
MIT

Code for a paper exploring using diffusion models to defend neural networks against adversarial attacks

101TeX, Jupyter NotebookDec 10, 20222 years agoMIT
adversarial-machine-learningdiffusion-modelsmachine-learningresnet

Similar Papers

DIFFender: Diffusion-Based Adversarial Defense against Patch Attacks

88% match1 repo382023

Real-world Adversarial Defense against Patch Attacks based on Diffusion Model

88% match2 repos382024

Robust Classification via a Single Diffusion Model

87% match2 repos972023

Struggle with Adversarial Defense? Try Diffusion

87% match2024

Diffusion or Non-Diffusion Adversarial Defenses: Rethinking the Relation between Classifier and Adversarial Purifier

87% match2025

A Low-Rank Defense Method for Adversarial Attack on Diffusion Models

87% match2026

Cite this paper

@article{silva2023diffdefense,
  title  = {DiffDefense: Defending against Adversarial Attacks via Diffusion Models},
  author = {Hondamunige Prasanna Silva and Lorenzo Seidenari and Alberto Del Bimbo},
  year   = {2023},
  eprint = {2309.03702},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2309.03702v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2309.03702.svg)](https://codeofpaper.com/paper/2309.03702)

Discussion