OptMark: Robust Multi-bit Diffusion Watermarking via Inference Time Optimization

2025-08-29

AI Summary

Diffusion-based image generation poses two key challenges for watermarking: zero-bit schemes lack scalability for large-scale user tracing, while multi-bit approaches suffer from insufficient robustness against generative attacks and common image transformations. This paper proposes an inference-time optimization framework for robust multi-bit watermarking. It is the first to jointly embed structured and fine-grained watermarks into intermediate latent spaces during the denoising process. We introduce an adjoint-gradient method that reduces memory complexity of optimization from O(N) to O(1). Additionally, we design a customized regularization term to enhance watermark imperceptibility and extraction stability without compromising visual fidelity. Experiments demonstrate that our method achieves over 92% bit extraction accuracy under diverse attacks—including numerical perturbations, geometric distortions, image editing, and regeneration—significantly outperforming state-of-the-art alternatives.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Search and Optimization: Distributed Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods

Abstract

Watermarking diffusion-generated images is crucial for copyright protection and user tracking. However, current diffusion watermarking methods face significant limitations: zero-bit watermarking systems lack the capacity for large-scale user tracking, while multi-bit methods are highly sensitive to certain image transformations or generative attacks, resulting in a lack of comprehensive robustness. In this paper, we propose OptMark, an optimization-based approach that embeds a robust multi-bit watermark into the intermediate latents of the diffusion denoising process. OptMark strategically inserts a structural watermark early to resist generative attacks and a detail watermark late to withstand image transformations, with tailored regularization terms to preserve image quality and ensure imperceptibility. To address the challenge of memory consumption growing linearly with the number of denoising steps during optimization, OptMark incorporates adjoint gradient methods, reducing memory usage from O(N) to O(1). Experimental results demonstrate that OptMark achieves invisible multi-bit watermarking while ensuring robust resilience against valuemetric transformations, geometric transformations, editing, and regeneration attacks.

Problem

• Robust multi-bit watermarking for diffusion-generated images
• Overcoming sensitivity to image transformations and generative attacks
• Reducing memory consumption during watermark optimization process

Innovation

• Optimization-based multi-bit watermark embedding in diffusion latents
• Adjoint gradient methods reduce memory from O(N) to O(1)
• Strategic early structural and late detail watermark insertion