tin-z / TimbreWatermarking

🐰 Detecting Voice Cloning Attacks via Timbre Watermarking

Source code for paper “Detecting Voice Cloning Attacks via Timbre Watermarking”

by Chang Liu, Jie Zhang, Tianwei Zhang, Xi Yang, Weiming Zhang, and Nenghai Yu In Network and Distributed System Security Symposium (NDSS) 2024.

Visit our website for audio samples.

Introduction

• watermarking_model: Code of the watermarking model
• voice.clone: Code and details of the voice cloning part

🐇 In this repository, we provide the complete code for training and testing the watermarking model. Additionally, we include the source code used for voice cloning experiments under various scenarios, along with corresponding README files. Please visit the respective directories to access detailed READMEs.

Model files

All the parameter files for the voice cloning model used in our work are available at this link.

Acknowledgments

Part of our experiments were based on code from several open-source repositories, including VITS, Tacotron2, so-vits-svc, Hifi-GAN, and FastSpeech2. Their code served as a foundation for portions of our experiments.

Citation

If you find this work useful, please consider citing our paper:

@inproceedings{timbrewatermarking-ndss2024,
  title = {Detecting Voice Cloning Attacks via Timbre Watermarking},
  author = {Liu, Chang and Zhang, Jie and Zhang, Tianwei and Yang, Xi and Zhang, Weiming and Yu, Nenghai},
  booktitle = {Network and Distributed System Security Symposium},
  year = {2024},
  doi = {10.14722/ndss.2024.24200},
}