Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multiplexing Neural Audio Watermarks

Zheqi Yuan, Yucheng Huang, Guangzhi Sun, Zengrui Jin, Chao Zhang

TL;DR

Experimental results on the LibriSpeech and Common Voice datasets under 14 diverse attack types demonstrate that both PA-TFM and MaskNet considerably outperform existing single-watermark baselines, establishing a resilient paradigm for real-world audio protection.

Abstract

Audio watermarking is essential for verifying speech authenticity, yet single-watermark schemes often struggle against sophisticated distortions such as neural reconstruction and adversarial attacks. To address this limitation, we introduce a multiplexing paradigm that combines multiple watermarking techniques to leverage their inherent complementarities. We explore both parallel and sequential multiplexing strategies and propose perceptual-adaptive time-frequency multiplexing (PA-TFM), a robust training-free approach. To further enhance performance, we introduce MaskNet, a novel model-based framework designed to learn effective time-domain multiplexing. Experimental results on the LibriSpeech and Common Voice datasets under 14 diverse attack types, including high-strength white-box and neural reconstruction attacks, demonstrate that both PA-TFM and MaskNet considerably outperform existing single-watermark baselines, establishing a resilient paradigm for real-world audio protection.

Multiplexing Neural Audio Watermarks

TL;DR

Experimental results on the LibriSpeech and Common Voice datasets under 14 diverse attack types demonstrate that both PA-TFM and MaskNet considerably outperform existing single-watermark baselines, establishing a resilient paradigm for real-world audio protection.

Abstract

Audio watermarking is essential for verifying speech authenticity, yet single-watermark schemes often struggle against sophisticated distortions such as neural reconstruction and adversarial attacks. To address this limitation, we introduce a multiplexing paradigm that combines multiple watermarking techniques to leverage their inherent complementarities. We explore both parallel and sequential multiplexing strategies and propose perceptual-adaptive time-frequency multiplexing (PA-TFM), a robust training-free approach. To further enhance performance, we introduce MaskNet, a novel model-based framework designed to learn effective time-domain multiplexing. Experimental results on the LibriSpeech and Common Voice datasets under 14 diverse attack types, including high-strength white-box and neural reconstruction attacks, demonstrate that both PA-TFM and MaskNet considerably outperform existing single-watermark baselines, establishing a resilient paradigm for real-world audio protection.

Paper Structure

This paper contains 14 sections, 6 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Parallel and Sequential Multiplexing
  5. Perceptual-Adaptive Time-Frequency Multiplexing
  6. MaskNet: Neural Time-Domain Fusion
  7. Experimental Setup
  8. Experimental Results
  9. Performance Evaluation and Generalization
  10. Exploration of Advanced and Extreme Threats
  11. Complementary Effects
  12. Acoustic Fidelity and Downstream Integrity
  13. Conclusion
  14. Generative AI Use Disclosure

Figures (2)

  • Figure 1: Overview of the proposed multiplexing architectures: (a) PA-TFM using heuristic time-frequency masking; (b) MaskNet uses learned time-domain masking. MaskNet optimises the mask distribution via a differentiable attack loop while keeping the pre-trained watermark extractors frozen.
  • Figure 2: TPR curves for watermark A and P under different attack SNR strengths. (a) Gaussian noise, where the watermark P degrades more slowly; (b) Room impulse response, where watermark A degrades more slowly. These complementary effects illustrate the benefit of combining both watermarks.