Interpretable Temporal Class Activation Representation for Audio Spoofing Detection

TL;DR

This paper introduces an interpretable audio spoofing detection model using wav2vec 2.0 and class activation representations, achieving state-of-the-art performance by localizing discriminative frames and learning attack-specific features.

Contribution

It integrates interpretability directly into the model architecture and employs multi-label training on spoofing types to improve detection accuracy.

Findings

Achieved EER of 0.51% on ASVspoof2019-LA

Developed class activation representation for frame localization

Enhanced detection performance with multi-label spoofing training

Abstract

Explaining the decisions made by audio spoofing detection models is crucial for fostering trust in detection outcomes. However, current research on the interpretability of detection models is limited to applying XAI tools to post-trained models. In this paper, we utilize the wav2vec 2.0 model and attentive utterance-level features to integrate interpretability directly into the model's architecture, thereby enhancing transparency of the decision-making process. Specifically, we propose a class activation representation to localize the discriminative frames contributing to detection. Furthermore, we demonstrate that multi-label training based on spoofing types, rather than binary labels as bonafide and spoofed, enables the model to learn distinct characteristics of different attacks, significantly improving detection performance. Our model achieves state-of-the-art results, with an EER of 0.51% and a min t-DCF of 0.0165 on the ASVspoof2019-LA set.