analiza aiqbench
This paper proposes a training-free defense for semantic watermarks in diffusion models, targeting attacks that exploit diffusion inversion to remove watermarks or forge them. Its core claim is that these attacks work by displacing latents across the watermark boundary, and its method tries to project them back. Readers should treat the results as preprint findings, since the reported gains depend on the specific watermarking schemes and attacks tested.
PGID: Progressive Guided Inversion and Denoising for Robust Watermark Detection
arXiv:2605.09319v2 Announce Type: replace-cross Abstract: With the proliferation of AI-generated images, digital watermarking has become an essential safeguard for protecting intellectual property and mitigating malicious exploitation. Recent works on semantic watermarking have enabled efficient copyright protection for diffusion models. However, the dependence of semantic watermarking on diffusion inversion for watermark detection creates a critical vulnerability. Imprint removal and forgery attacks exploit this weakness to produce deceptive results. Our analysis reveals that these attacks succeed by displacing watermarked latents into the unwatermarked region, while guiding unwatermarked latents into the watermarked region. Based on that, we propose Progressive Guided Inversion and Denoising (PGID), the first plug-and-play, training-free noise extraction framework designed to defend against both attack strategies. PGID effectively defends by projecting perturbed latents back to the region where they originally belong. The projection is achieved by eliminating intermediate latent deflections and mitigating adversarial perturbations through progressive inversion-denoising cycles. Comprehensive evaluations across multiple schemes demonstrate that PGID successfully restores detection reliability by recovering removed watermarks and identifying forged instances.
Abstract:With the proliferation of AI-generated images, digital watermarking has become an essential safeguard for protecting intellectual property and mitigating malicious exploitation. Recent works on semantic watermarking have enabled efficient copyright protection for diffusion models. However, the dependence of semantic watermarking on diffusion inversion for watermark detection creates a critical vulnerability. Imprint removal and forgery attacks exploit this weakness to produce deceptive results. Our analysis reveals that these attacks succeed by displacing watermarked latents into the unwatermarked region, while guiding unwatermarked latents into the watermarked region. Based on that, we propose Progressive Guided Inversion and Denoising (PGID), the first plug-and-play, training-free noise extraction framework designed to defend against both attack strategies. PGID effectively defends by projecting perturbed latents back to the region where they originally belong. The projection is achieved by eliminating intermediate latent deflections and mitigating adversarial perturbations through progressive inversion-denoising cycles. Comprehensive evaluations across multiple schemes demonstrate that PGID successfully restores detection reliability by recovering removed watermarks and identifying forged instances.
Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG)
arXiv:2605.09319 [cs.CV]
(or arXiv:2605.09319v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2605.09319 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 260 cuvinte
Subiecte conexe
DenMark: Robust Semantic Watermarking for Diffusion Language Models
arXiv · 15 sept. 2026
Build a Physical AI model factory with NVIDIA Cosmos 3 on SageMaker HyperPod
AWS · 4 sept. 2026
RadioVIL: Anomaly-Aware Diffusion Models for Radio Map Inpainting and Zero-Shot Vehicle Localization
arXiv · 19 aug. 2026
Informe de riesgos de la IA de frontera (febrero–marzo de 2026)
METR · 19 mai 2026