CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models

Sen Peng, Mingyue Wang, Jianfei He, Jijia Yang, Xiaohua Jia
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:48872-48885, 2025.

Abstract

Latent diffusion models have recently demonstrated superior capabilities in many downstream image synthesis tasks. However, customization of latent diffusion models using unauthorized data can severely compromise the privacy and intellectual property rights of data owners. Adversarial examples as protective perturbations have been developed to defend against unauthorized data usage by introducing imperceptible noise to customization samples, preventing diffusion models from effectively learning them. In this paper, we first reveal that the primary reason adversarial examples are effective as protective perturbations in latent diffusion models is the distortion of their latent representations, as demonstrated through qualitative and quantitative experiments. We then propose the Contrastive Adversarial Training (CAT) utilizing lightweight adapters as an adaptive attack against these protection methods, highlighting their lack of robustness. Extensive experiments demonstrate that our CAT method significantly reduces the effectiveness of protective perturbations in customization, urging the community to reconsider and improve the robustness of existing protective perturbations. The code is available at https://github.com/senp98/CAT.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-peng25e, title = {{CAT}: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models}, author = {Peng, Sen and Wang, Mingyue and He, Jianfei and Yang, Jijia and Jia, Xiaohua}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {48872--48885}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/peng25e/peng25e.pdf}, url = {https://proceedings.mlr.press/v267/peng25e.html}, abstract = {Latent diffusion models have recently demonstrated superior capabilities in many downstream image synthesis tasks. However, customization of latent diffusion models using unauthorized data can severely compromise the privacy and intellectual property rights of data owners. Adversarial examples as protective perturbations have been developed to defend against unauthorized data usage by introducing imperceptible noise to customization samples, preventing diffusion models from effectively learning them. In this paper, we first reveal that the primary reason adversarial examples are effective as protective perturbations in latent diffusion models is the distortion of their latent representations, as demonstrated through qualitative and quantitative experiments. We then propose the Contrastive Adversarial Training (CAT) utilizing lightweight adapters as an adaptive attack against these protection methods, highlighting their lack of robustness. Extensive experiments demonstrate that our CAT method significantly reduces the effectiveness of protective perturbations in customization, urging the community to reconsider and improve the robustness of existing protective perturbations. The code is available at https://github.com/senp98/CAT.} }
Endnote
%0 Conference Paper %T CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models %A Sen Peng %A Mingyue Wang %A Jianfei He %A Jijia Yang %A Xiaohua Jia %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-peng25e %I PMLR %P 48872--48885 %U https://proceedings.mlr.press/v267/peng25e.html %V 267 %X Latent diffusion models have recently demonstrated superior capabilities in many downstream image synthesis tasks. However, customization of latent diffusion models using unauthorized data can severely compromise the privacy and intellectual property rights of data owners. Adversarial examples as protective perturbations have been developed to defend against unauthorized data usage by introducing imperceptible noise to customization samples, preventing diffusion models from effectively learning them. In this paper, we first reveal that the primary reason adversarial examples are effective as protective perturbations in latent diffusion models is the distortion of their latent representations, as demonstrated through qualitative and quantitative experiments. We then propose the Contrastive Adversarial Training (CAT) utilizing lightweight adapters as an adaptive attack against these protection methods, highlighting their lack of robustness. Extensive experiments demonstrate that our CAT method significantly reduces the effectiveness of protective perturbations in customization, urging the community to reconsider and improve the robustness of existing protective perturbations. The code is available at https://github.com/senp98/CAT.
APA
Peng, S., Wang, M., He, J., Yang, J. & Jia, X.. (2025). CAT: Contrastive Adversarial Training for Evaluating the Robustness of Protective Perturbations in Latent Diffusion Models. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:48872-48885 Available from https://proceedings.mlr.press/v267/peng25e.html.

Related Material