On the Coexistence and Ensembling of Watermarks

2025-01-29

AI Summary

Deep image watermarking methods are inherently mutually exclusive, making concurrent embedding and decoding of multiple watermarks infeasible. Method: We propose the first integrated framework enabling cooperative multi-watermark embedding and parallel decoding. Leveraging mainstream architectures (e.g., U-Net), it employs non-intrusive spatial/frequency-domain superposition embedding and independent decoding—requiring no retraining for compatibility with diverse watermark schemes. Contribution/Results: This work is the first to systematically validate the coexistence feasibility of open-source deep watermarks and establishes a controllable trade-off paradigm among capacity, robustness, fidelity, and decoding accuracy. Experiments show that under multi-watermark coexistence: PSNR degradation remains below 0.5 dB; average decoding accuracy exceeds 92%; aggregate message capacity increases by 3.2×; and robustness improves rather than degrades. These results overcome fundamental limitations of conventional single-task watermarking, enabling high-density, multi-purpose, and highly stable image watermarking applications.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Watermarking, the practice of embedding imperceptible information into media such as images, videos, audio, and text, is essential for intellectual property protection, content provenance and attribution. The growing complexity of digital ecosystems necessitates watermarks for different uses to be embedded in the same media. However, to detect and decode all watermarks, they need to coexist well with one another. We perform the first study of coexistence of deep image watermarking methods and, contrary to intuition, we find that various open-source watermarks can coexist with only minor impacts on image quality and decoding robustness. The coexistence of watermarks also opens the avenue for ensembling watermarking methods. We show how ensembling can increase the overall message capacity and enable new trade-offs between capacity, accuracy, robustness and image quality, without needing to retrain the base models.

Problem

• Watermark Compatibility
• Image Clarity
• Information Embedding

Innovation

• Advanced Watermarking Techniques
• Synergy and Combination
• Enhanced Information Embedding