SPFM-Net: Semantic-Prior-Guided Frequency-Constrained Mamba for Invisible Watermark Attack

2026-07-30

AI Summary

Existing watermark removal methods struggle to effectively model the long-range dependencies of global watermark signals, often failing to balance removal efficacy with visual fidelity. This work proposes a Mamba-based architecture that integrates semantic priors and frequency-domain constraints to address this challenge. By applying high-ratio masking to disrupt the spatial coherence of watermarks and leveraging a partially fine-tuned masked autoencoder for semantically consistent image reconstruction, the method enhances content preservation. Furthermore, it introduces multi-scale residual frequency-domain feature interaction and Mamba-driven global state-space modeling to jointly optimize spatial, spectral, and edge information. To the best of our knowledge, this is the first approach to embed semantic guidance and frequency-domain constraints within a Mamba framework, achieving state-of-the-art attack performance across multiple mainstream watermarking schemes while maintaining superior visual quality and striking an optimal balance between watermark suppression and image fidelity.

Abstract

Existing watermark attacks typically rely on predefined signal-processing operations or locally constrained restoration networks, making it difficult to capture the long-range dependencies of globally distributed watermark signals and resulting in an unfavorable trade-off between removal effectiveness and visual fidelity. In this paper, we propose SPFM-Net, a semantic-prior-guided and frequency-constrained Mamba framework for invisible watermark attack. SPFM-Net first employs high-ratio masking to disrupt the spatial coherence of invisible watermark signals, and then utilizes a partially fine-tuned pretrained Masked Autoencoder to reconstruct semantically consistent image from sparse observations while suppressing watermark-related information. A Multi-scale Residual Frequency Feature Interaction module subsequently aggregates watermark-related residual features across multiple receptive fields, while adaptively suppressing responses from watermark-irrelevant regions. To further capture the long-range dependencies of globally distributed watermark signals, a lightweight Mamba-based Global State-space Feature Modeling (GSFM) unit is introduced to separate watermark-related features from natural image content and suppress the remaining watermark traces. In addition, SPFM-Net is optimized using a multi-level objective that jointly imposes spatial-, frequency-, and edge-domain constraints, enabling effective watermark suppression while preserving perceptual quality. Extensive experiments on representative spatial-domain, transform-domain, orthogonal moment-based, and deep learning-based watermarking schemes demonstrate that SPFM-Net achieves a favorable trade-off between watermark attack effectiveness and perceptual fidelity.

Problem

• invisible watermark attack
• long-range dependencies
• visual fidelity
• watermark removal
• global watermark signals

Innovation

• Mamba
• frequency-constrained
• semantic prior
• invisible watermark attack
• global state-space modeling