VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Chou, Sheng-Yen, Chen, Pin-Yu, Ho, Tsung-Yi
Format: Preprint
Publié: 2023
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909056410583040
author Chou, Sheng-Yen
Chen, Pin-Yu
Ho, Tsung-Yi
author_facet Chou, Sheng-Yen
Chen, Pin-Yu
Ho, Tsung-Yi
contents Diffusion Models (DMs) are state-of-the-art generative models that learn a reversible corruption process from iterative noise addition and denoising. They are the backbone of many generative AI applications, such as text-to-image conditional generation. However, recent studies have shown that basic unconditional DMs (e.g., DDPM and DDIM) are vulnerable to backdoor injection, a type of output manipulation attack triggered by a maliciously embedded pattern at model input. This paper presents a unified backdoor attack framework (VillanDiffusion) to expand the current scope of backdoor analysis for DMs. Our framework covers mainstream unconditional and conditional DMs (denoising-based and score-based) and various training-free samplers for holistic evaluations. Experiments show that our unified framework facilitates the backdoor analysis of different DM configurations and provides new insights into caption-based backdoor attacks on DMs. Our code is available on GitHub: \url{https://github.com/IBM/villandiffusion}
format Preprint
id arxiv_https___arxiv_org_abs_2306_06874
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models
Chou, Sheng-Yen
Chen, Pin-Yu
Ho, Tsung-Yi
Cryptography and Security
Computer Vision and Pattern Recognition
Machine Learning
Diffusion Models (DMs) are state-of-the-art generative models that learn a reversible corruption process from iterative noise addition and denoising. They are the backbone of many generative AI applications, such as text-to-image conditional generation. However, recent studies have shown that basic unconditional DMs (e.g., DDPM and DDIM) are vulnerable to backdoor injection, a type of output manipulation attack triggered by a maliciously embedded pattern at model input. This paper presents a unified backdoor attack framework (VillanDiffusion) to expand the current scope of backdoor analysis for DMs. Our framework covers mainstream unconditional and conditional DMs (denoising-based and score-based) and various training-free samplers for holistic evaluations. Experiments show that our unified framework facilitates the backdoor analysis of different DM configurations and provides new insights into caption-based backdoor attacks on DMs. Our code is available on GitHub: \url{https://github.com/IBM/villandiffusion}
title VillanDiffusion: A Unified Backdoor Attack Framework for Diffusion Models
topic Cryptography and Security
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2306.06874