Shallow Diffuse: Robust and Invisible Watermarking through Low-Dimensional Subspaces in Diffusion Models

2024-10-28

arXiv.org

AI Summary

To address copyright infringement and misinformation arising from AI-generated images, this paper proposes a robust, imperceptible watermarking method tailored for diffusion model outputs. The core innovation is a novel decoupled embedding mechanism: leveraging geometric analysis in the latent space, we construct a low-rank subspace and project the watermark into its null space—thereby achieving mathematical decoupling between the watermark and semantic image content. This design ensures both perceptual invisibility and statistical detectability. Extensive experiments demonstrate that our method achieves a 12.6% higher watermark detection rate under common attacks compared to state-of-the-art approaches, while preserving the Fréchet Inception Distance (FID) at zero degradation. These results significantly outperform existing watermarking schemes, offering an efficient and reliable foundation for copyright attribution and forgery prevention in generative AI content.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Generation

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

The widespread use of AI-generated content from diffusion models has raised significant concerns regarding misinformation and copyright infringement. Watermarking is a crucial technique for identifying these AI-generated images and preventing their misuse. In this paper, we introduce Shallow Diffuse, a new watermarking technique that embeds robust and invisible watermarks into diffusion model outputs. Unlike existing approaches that integrate watermarking throughout the entire diffusion sampling process, Shallow Diffuse decouples these steps by leveraging the presence of a low-dimensional subspace in the image generation process. This method ensures that a substantial portion of the watermark lies in the null space of this subspace, effectively separating it from the image generation process. Our theoretical and empirical analyses show that this decoupling strategy greatly enhances the consistency of data generation and the detectability of the watermark. Extensive experiments further validate that our Shallow Diffuse outperforms existing watermarking methods in terms of robustness and consistency. The codes will be released at https://github.com/liwd190019/Shallow-Diffuse.

Problem

• Robust invisible watermarking for diffusion model outputs
• Decoupling watermarking from image generation via low-dimensional subspaces
• Enhancing watermark detectability and generation consistency

Innovation

• Leverages low-dimensional subspaces for watermarking
• Decouples watermarking from diffusion sampling process
• Enhances watermark robustness and detectability