Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift
Abstract
Diffusion models (DM) have become state-of-the-art generative models because of their capability of generating high-quality images from noises without adversarial training. However, they are vulnerable to backdoor attacks as reported by recent studies. When a data input (e.g., some Gaussian noise) is stamped with a trigger (e.g., a white patch), the backdoored model always generates the target image (e.g., an improper photo). However, effective defense strategies to mitigate backdoors from DMs are underexplored. To bridge this gap, we propose the first backdoor detection and removal framework for DMs. We evaluate our framework Elijah on over hundreds of DMs of 3 types including DDPM, NCSN and LDM, with 13 samplers against 3 existing backdoor attacks. Extensive experiments show that our approach can have close to 100% detection accuracy and reduce the backdoor effects to close to zero without significantly sacrificing the model utility.
Cite
Text
An et al. "Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift." AAAI Conference on Artificial Intelligence, 2024. doi:10.1609/AAAI.V38I10.28958Markdown
[An et al. "Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift." AAAI Conference on Artificial Intelligence, 2024.](https://mlanthology.org/aaai/2024/an2024aaai-elijah/) doi:10.1609/AAAI.V38I10.28958BibTeX
@inproceedings{an2024aaai-elijah,
title = {{Elijah: Eliminating Backdoors Injected in Diffusion Models via Distribution Shift}},
author = {An, Shengwei and Chou, Sheng-Yen and Zhang, Kaiyuan and Xu, Qiuling and Tao, Guanhong and Shen, Guangyu and Cheng, Siyuan and Ma, Shiqing and Chen, Pin-Yu and Ho, Tsung-Yi and Zhang, Xiangyu},
booktitle = {AAAI Conference on Artificial Intelligence},
year = {2024},
pages = {10847-10855},
doi = {10.1609/AAAI.V38I10.28958},
url = {https://mlanthology.org/aaai/2024/an2024aaai-elijah/}
}