We Can Hide More Bits: The Unused Watermarking Capacity in Theory and in Practice

2025-10-07

AI Summary

Current deep learning–based image watermarking methods achieve only hundreds of bits of payload capacity—far below the theoretical upper bound—indicating substantial room for fundamental improvement. Method: We first derive, under PSNR and linear robustness constraints, a rigorous theoretical upper bound on watermark capacity for images; analysis reveals that existing models exploit only a tiny fraction of this bound. Guided by an architectural scalability hypothesis, we propose ChunkySeal—a lightweight, efficient model inspired by VideoSeal—that employs a block-wise embedding architecture and tailored training strategies. Contribution/Results: ChunkySeal achieves 1024-bit capacity while maintaining high visual fidelity (PSNR ≥ 42 dB) and robustness—quadrupling the capacity of state-of-the-art methods. This work demonstrates the practical feasibility of high-capacity watermarking and establishes a “capacity-driven architecture design” paradigm, providing both theoretical foundations and technical pathways for modern steganography and generative watermarking.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Deep Generative Models & Autoencoders
• Search and Optimization: Learning to Search

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Efficiency and scalability of Web search engines
• Responsible Web: Measurement, analysis, and circumvention of Web censorship

Abstract

Despite rapid progress in deep learning-based image watermarking, the capacity of current robust methods remains limited to the scale of only a few hundred bits. Such plateauing progress raises the question: How far are we from the fundamental limits of image watermarking? To this end, we present an analysis that establishes upper bounds on the message-carrying capacity of images under PSNR and linear robustness constraints. Our results indicate theoretical capacities are orders of magnitude larger than what current models achieve. Our experiments show this gap between theoretical and empirical performance persists, even in minimal, easily analysable setups. This suggests a fundamental problem. As proof that larger capacities are indeed possible, we train ChunkySeal, a scaled-up version of VideoSeal, which increases capacity 4 times to 1024 bits, all while preserving image quality and robustness. These findings demonstrate modern methods have not yet saturated watermarking capacity, and that significant opportunities for architectural innovation and training strategies remain.

Problem

• Establishes theoretical upper bounds for image watermarking capacity
• Identifies large gap between theoretical and current empirical capacities
• Demonstrates potential for significantly higher robust watermarking capacity

Innovation

• Establishes theoretical watermarking capacity upper bounds
• Proposes scaled-up ChunkySeal model architecture
• Achieves 1024-bit capacity while maintaining robustness