Automatic Speaker Verification Spoofing and Deepfake Detection Using Wav2vec 2.0 and Data Augmentation

The performance of spoofing countermeasure systems depends fundamentally upon use of sufficiently representative training data. With this usually being, current solutions typically lack generalisation to attacks encountered the wild. Strategies to improve reliability in the face of uncontrolled, attacks are hence needed. We report in this paper our efforts to self-supervised learning in the form of a wav2vec 2. 0 front-end with fine. Despite initial base representations being learned using only bona fide and no spoofed data, we obtain the lowest equal error rates reported in literature for both the ASVspoof 2021 Logical Access and Deepfake. When combined with data augmentation, these results correspond to an of almost 90% relative to our baseline system.