Collaborative Watermarking for Adversarial Speech Synthesis

TL;DR

This paper introduces a collaborative training approach for watermarking neural speech synthesis, enhancing detection accuracy while maintaining speech quality, and demonstrating robustness against noise and distortions.

Contribution

It proposes a novel collaborative training scheme for speech watermarking that improves detection and robustness without degrading perceptual quality.

Findings

Collaborative training improves detection performance over traditional methods.

The approach enhances robustness against noise and time-stretching.

Listening tests show minimal impact on speech quality.

Abstract

Advances in neural speech synthesis have brought us technology that is not only close to human naturalness, but is also capable of instant voice cloning with little data, and is highly accessible with pre-trained models available. Naturally, the potential flood of generated content raises the need for synthetic speech detection and watermarking. Recently, considerable research effort in synthetic speech detection has been related to the Automatic Speaker Verification and Spoofing Countermeasure Challenge (ASVspoof), which focuses on passive countermeasures. This paper takes a complementary view to generated speech detection: a synthesis system should make an active effort to watermark the generated speech in a way that aids detection by another machine, but remains transparent to a human listener. We propose a collaborative training scheme for synthetic speech watermarking and show that a HiFi-GAN neural vocoder collaborating with the ASVspoof 2021 baseline countermeasure models consistently improves detection performance over conventional classifier training. Furthermore, we demonstrate how collaborative training can be paired with augmentation strategies for added robustness against noise and time-stretching. Finally, listening tests demonstrate that collaborative training has little adverse effect on perceptual quality of vocoded speech.