Continual Learning for Fake Audio Detection

TL;DR

This paper introduces a continual learning approach for fake audio detection that effectively learns new spoofing attacks without forgetting previous knowledge, outperforming traditional fine-tuning methods.

Contribution

It proposes a novel continual-learning-based method with knowledge distillation and embedding similarity loss to improve fake audio detection across unseen attacks.

Findings

Outperforms fine-tuning with up to 81.62% EER reduction

Effective in incremental learning of new spoofing attacks

Maintains performance on previous data without retraining

Abstract

Fake audio attack becomes a major threat to the speaker verification system. Although current detection approaches have achieved promising results on dataset-specific scenarios, they encounter difficulties on unseen spoofing data. Fine-tuning and retraining from scratch have been applied to incorporate new data. However, fine-tuning leads to performance degradation on previous data. Retraining takes a lot of time and computation resources. Besides, previous data are unavailable due to privacy in some situations. To solve the above problems, this paper proposes detecting fake without forgetting, a continual-learning-based method, to make the model learn new spoofing attacks incrementally. A knowledge distillation loss is introduced to loss function to preserve the memory of original model. Supposing the distribution of genuine voice is consistent among different scenarios, an extra embedding similarity loss is used as another constraint to further do a positive sample alignment. Experiments are conducted on the ASVspoof2019 dataset. The results show that our proposed method outperforms fine-tuning by the relative reduction of average equal error rate up to 81.62%.