Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Automatic speaker verification spoofing and deepfake detection using wav2vec 2.0 and data augmentation

About

The performance of spoofing countermeasure systems depends fundamentally upon the use of sufficiently representative training data. With this usually being limited, current solutions typically lack generalisation to attacks encountered in the wild. Strategies to improve reliability in the face of uncontrolled, unpredictable attacks are hence needed. We report in this paper our efforts to use self-supervised learning in the form of a wav2vec 2.0 front-end with fine tuning. Despite initial base representations being learned using only bona fide data and no spoofed data, we obtain the lowest equal error rates reported in the literature for both the ASVspoof 2021 Logical Access and Deepfake databases. When combined with data augmentation,these results correspond to an improvement of almost 90% relative to our baseline system.

Hemlata Tak, Massimiliano Todisco, Xin Wang, Jee-weon Jung, Junichi Yamagishi, Nicholas Evans• 2022

Related benchmarks

TaskDatasetResultRank
Audio Deepfake Detectionin the wild
EER7.58
58
Spoof Speech DetectionASVspoof LA 2021 (eval)
min-tDCF0.212
36
Audio Deepfake DetectionASVspoof DF 2021
EER4.08
35
Audio Deepfake DetectionASVspoof 2021
EER3.6
27
Audio Deepfake DetectionASVspoof 2019
EER0.2
25
Speech Spoofing DetectionIn-the-Wild (ITW) (eval)
EER2.09
19
Synthetic Speech DetectionASVspoof DF 2021 (eval)
EER (%)3.69
19
Audio Deepfake DetectionMLAAD-EN
EER14.5
18
Audio Deepfake DetectionASVspoof LA and DF 2021
EER (DF)2.85
17
Audio Deepfake DetectionASVspoof LA 2021
EER0.82
12
Showing 10 of 37 rows

Other info

Follow for update