# Transferable Multi-Bit Watermarking Across Frozen Diffusion Models via Latent Consistency Bridges

Source: [Transferable Multi-Bit Watermarking Across Frozen Diffusion Models via Latent Consistency Bridges](<https://www.sota2.com/research/paper/69c21a880636e5bf4afc480c>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-05T05:00:04.417Z

Authors: Hong-Hanh Nguyen-Le, Van-Tuan Tran, Thuc D. Nguyen, Nhien-An Le-Khac

Published: 2026-03-19T13:13:34.000Z

## Abstract

As generative AI advances, global governance frameworks increasingly mandate verifiable content provenance. However, existing watermarking techniques face a critical policy-to-technology disconnect: sampling-based methods require computationally prohibitive inversion, while fine-tuning approaches are tethered to specific model checkpoints, hindering standardized, cross-model oversight. To bridge this gap, we introduce DiffMark, a plug-and-play multi-bit watermarking framework. DiffMark embeds a persistent, learned perturbation into every denoising step of a frozen diffusion model, accumulating a recoverable signal in the final latent space. To enable efficient training through the frozen network, we utilize Latent Consistency Models (LCMs) as a differentiable training bridge. DiffMark achieves 64-bit extraction in a single 16.4 ms forward pass, which is a $45\\times$ speed-up over inversion baselines. By enabling per-image key flexibility and cross-architecture transferability without retraining, DiffMark provides the practical, scalable technical tooling necessary to operationalize user accountability and enforce emerging AI governance mandates.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Image Watermarking on DALL-E 3](<https://www.sota2.com/research/sota/image-watermarking-on-dall-e-3.md>): Image Watermarking; DALL-E 3; PSNR: 9.75.

- [Watermark Robustness on DiffusionDB](<https://www.sota2.com/research/sota/watermark-robustness-on-diffusiondb.md>): Watermark Robustness; DiffusionDB; Brightness Robustness: 100.

- [Watermark Robustness on DALL-E 3](<https://www.sota2.com/research/sota/watermark-robustness-on-dall-e-3.md>): Watermark Robustness; DALL-E 3; Robustness: Brightness: 100.

- [Image Watermarking on MS-COCO](<https://www.sota2.com/research/sota/image-watermarking-on-ms-coco.md>): Image Watermarking; MS-COCO; PSNR: 8.91.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2603.20304>)

[PDF](<https://arxiv.org/pdf/2603.20304.pdf>)