AquaLoRA: Toward White-box Protection for Customized Stable Diffusion Models via Watermark LoRA

TL;DR

AquaLoRA introduces a white-box watermarking method for customized Stable Diffusion models by embedding watermarks via a LoRA module, ensuring robustness against removal or replacement attempts while maintaining high image fidelity.

Contribution

The paper presents AquaLoRA, the first approach enabling white-box watermark protection in diffusion models through a novel LoRA-based watermark embedding and a fine-tuning strategy.

Findings

Effective watermark embedding with minimal impact on model performance.

Robustness against removal or tampering of watermarks.

Validated through extensive experiments and ablation studies.

Abstract

Diffusion models have achieved remarkable success in generating high-quality images. Recently, the open-source models represented by Stable Diffusion (SD) are thriving and are accessible for customization, giving rise to a vibrant community of creators and enthusiasts. However, the widespread availability of customized SD models has led to copyright concerns, like unauthorized model distribution and unconsented commercial use. To address it, recent works aim to let SD models output watermarked content for post-hoc forensics. Unfortunately, none of them can achieve the challenging white-box protection, wherein the malicious user can easily remove or replace the watermarking module to fail the subsequent verification. For this, we propose \texttt{\method} as the first implementation under this scenario. Briefly, we merge watermark information into the U-Net of Stable Diffusion Models via a watermark Low-Rank Adaptation (LoRA) module in a two-stage manner. For watermark LoRA module, we devise a scaling matrix to achieve flexible message updates without retraining. To guarantee fidelity, we design Prior Preserving Fine-Tuning (PPFT) to ensure watermark learning with minimal impacts on model distribution, validated by proofs. Finally, we conduct extensive experiments and ablation studies to verify our design.