Saved in:
Bibliographic Details
Main Authors: Wu, Chia-Hua, Ge, Wanying, Wang, Xin, Yamagishi, Junichi, Tsao, Yu, Wang, Hsin-Min
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2506.14398
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915348395065344
author Wu, Chia-Hua
Ge, Wanying
Wang, Xin
Yamagishi, Junichi
Tsao, Yu
Wang, Hsin-Min
author_facet Wu, Chia-Hua
Ge, Wanying
Wang, Xin
Yamagishi, Junichi
Tsao, Yu
Wang, Hsin-Min
contents Solutions for defending against deepfake speech fall into two categories: proactive watermarking models and passive conventional deepfake detectors. While both address common threats, their differences in training, optimization, and evaluation prevent a unified protocol for joint evaluation and selecting the best solutions for different cases. This work proposes a framework to evaluate both model types in deepfake speech detection. To ensure fair comparison and minimize discrepancies, all models were trained and tested on common datasets, with performance evaluated using a shared metric. We also analyze their robustness against various adversarial attacks, showing that different models exhibit distinct vulnerabilities to different speech attribute distortions. Our training and evaluation code is available at Github.
format Preprint
id arxiv_https___arxiv_org_abs_2506_14398
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle A Comparative Study on Proactive and Passive Detection of Deepfake Speech
Wu, Chia-Hua
Ge, Wanying
Wang, Xin
Yamagishi, Junichi
Tsao, Yu
Wang, Hsin-Min
Sound
Solutions for defending against deepfake speech fall into two categories: proactive watermarking models and passive conventional deepfake detectors. While both address common threats, their differences in training, optimization, and evaluation prevent a unified protocol for joint evaluation and selecting the best solutions for different cases. This work proposes a framework to evaluate both model types in deepfake speech detection. To ensure fair comparison and minimize discrepancies, all models were trained and tested on common datasets, with performance evaluated using a shared metric. We also analyze their robustness against various adversarial attacks, showing that different models exhibit distinct vulnerabilities to different speech attribute distortions. Our training and evaluation code is available at Github.
title A Comparative Study on Proactive and Passive Detection of Deepfake Speech
topic Sound
url https://arxiv.org/abs/2506.14398