BUT Systems and Analyses for the ASVspoof 5 Challenge

TL;DR

This paper presents the BUT systems for the ASVspoof 5 challenge, including deepfake detection models, analysis of label schemes, and a novel score fusion method for spoofing-robust speaker verification.

Contribution

It introduces ResNet18 and self-supervised models for detection, analyzes label schemes, and proposes a logistic regression-based score fusion for improved SASV performance.

Findings

ResNet18 and self-supervised models perform effectively in detection tasks.

Analyzing label schemes provides insights into training strategies.

Proposed score fusion improves spoofing-robust speaker verification results.

Abstract

This paper describes the BUT submitted systems for the ASVspoof 5 challenge, along with analyses. For the conventional deepfake detection task, we use ResNet18 and self-supervised models for the closed and open conditions, respectively. In addition, we analyze and visualize different combinations of speaker information and spoofing information as label schemes for training. For spoofing-robust automatic speaker verification (SASV), we introduce effective priors and propose using logistic regression to jointly train affine transformations of the countermeasure scores and the automatic speaker verification scores in such a way that the SASV LLR is optimized.