Towards Out-of-Distribution Detection in Vocoder Recognition via Latent Feature Reconstruction

TL;DR

This paper introduces a reconstruction-based out-of-distribution detection method for vocoder recognition using autoencoders, contrastive learning, and auxiliary classifiers, improving accuracy over existing probability-score methods.

Contribution

The study proposes a novel autoencoder-based OOD detection approach that employs contrastive learning and auxiliary classifiers to enhance detection accuracy in deepfake speech recognition.

Findings

Outperforms baseline systems by 10% in accuracy

Contrastive learning improves feature distinction

Auxiliary classifier enhances reconstruction quality

Abstract

Advancements in synthesized speech have created a growing threat of impersonation, making it crucial to develop deepfake algorithm recognition. One significant aspect is out-of-distribution (OOD) detection, which has gained notable attention due to its important role in deepfake algorithm recognition. However, most of the current approaches for detecting OOD in deepfake algorithm recognition rely on probability-score or classified-distance, which may lead to limitations in the accuracy of the sample at the edge of the threshold. In this study, we propose a reconstruction-based detection approach that employs an autoencoder architecture to compress and reconstruct the acoustic feature extracted from a pre-trained WavLM model. Each acoustic feature belonging to a specific vocoder class is only aptly reconstructed by its corresponding decoder. When none of the decoders can satisfactorily reconstruct a feature, it is classified as an OOD sample. To enhance the distinctiveness of the reconstructed features by each decoder, we incorporate contrastive learning and an auxiliary classifier to further constrain the reconstructed feature. Experiments demonstrate that our proposed approach surpasses baseline systems by a relative margin of 10\% in the evaluation dataset. Ablation studies further validate the effectiveness of both the contrastive constraint and the auxiliary classifier within our proposed approach.