SimuFreeMark: A Noise-Simulation-Free Robust Watermarking Against Image Editing
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866912709197430784 |
|---|---|
| author | Tang, Yichao Li, Mingyang Miao, Di Li, Sheng Qian, Zhenxing Zhang, Xinpeng |
| author_facet | Tang, Yichao Li, Mingyang Miao, Di Li, Sheng Qian, Zhenxing Zhang, Xinpeng |
| contents | The advancement of artificial intelligence generated content (AIGC) has created a pressing need for robust image watermarking that can withstand both conventional signal processing and novel semantic editing attacks. Current deep learning-based methods rely on training with hand-crafted noise simulation layers, which inherently limit their generalization to unforeseen distortions. In this work, we propose $\textbf{SimuFreeMark}$, a noise-$\underline{\text{simu}}$lation-$\underline{\text{free}}$ water$\underline{\text{mark}}$ing framework that circumvents this limitation by exploiting the inherent stability of image low-frequency components. We first systematically establish that low-frequency components exhibit significant robustness against a wide range of attacks. Building on this foundation, SimuFreeMark embeds watermarks directly into the deep feature space of the low-frequency components, leveraging a pre-trained variational autoencoder (VAE) to bind the watermark with structurally stable image representations. This design completely eliminates the need for noise simulation during training. Extensive experiments demonstrate that SimuFreeMark outperforms state-of-the-art methods across a wide range of conventional and semantic attacks, while maintaining superior visual quality. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2511_11295 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | SimuFreeMark: A Noise-Simulation-Free Robust Watermarking Against Image Editing Tang, Yichao Li, Mingyang Miao, Di Li, Sheng Qian, Zhenxing Zhang, Xinpeng Computer Vision and Pattern Recognition The advancement of artificial intelligence generated content (AIGC) has created a pressing need for robust image watermarking that can withstand both conventional signal processing and novel semantic editing attacks. Current deep learning-based methods rely on training with hand-crafted noise simulation layers, which inherently limit their generalization to unforeseen distortions. In this work, we propose $\textbf{SimuFreeMark}$, a noise-$\underline{\text{simu}}$lation-$\underline{\text{free}}$ water$\underline{\text{mark}}$ing framework that circumvents this limitation by exploiting the inherent stability of image low-frequency components. We first systematically establish that low-frequency components exhibit significant robustness against a wide range of attacks. Building on this foundation, SimuFreeMark embeds watermarks directly into the deep feature space of the low-frequency components, leveraging a pre-trained variational autoencoder (VAE) to bind the watermark with structurally stable image representations. This design completely eliminates the need for noise simulation during training. Extensive experiments demonstrate that SimuFreeMark outperforms state-of-the-art methods across a wide range of conventional and semantic attacks, while maintaining superior visual quality. |
| title | SimuFreeMark: A Noise-Simulation-Free Robust Watermarking Against Image Editing |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2511.11295 |