Gender Fairness in Audio Deepfake Detection: Performance and Disparity Analysis

TL;DR

This paper analyzes gender bias in audio deepfake detection models, revealing disparities hidden by traditional metrics and emphasizing the need for fairness-aware evaluation to improve system equity and robustness.

Contribution

It provides a comprehensive gender fairness analysis of audio deepfake detection models using multiple metrics and highlights the limitations of conventional performance measures.

Findings

Fairness metrics reveal gender disparities not shown by standard EER.

Gender bias persists even when overall performance appears balanced.

Fairness-aware evaluation is crucial for developing equitable detection systems.

Abstract

Audio deepfake detection aims to detect real human voices from those generated by Artificial Intelligence (AI) and has emerged as a significant problem in the field of voice biometrics systems. With the ever-improving quality of synthetic voice, the probability of such a voice being exploited for illicit practices like identity thest and impersonation increases. Although significant progress has been made in the field of Audio Deepfake Detection in recent times, the issue of gender bias remains underexplored and in its nascent stage In this paper, we have attempted a thorough analysis of gender dependent performance and fairness in audio deepfake detection models. We have used the ASVspoof 5 dataset and train a ResNet-18 classifier and evaluate detection performance across four different audio features, and compared the performance with baseline AASIST model. Beyond conventional metrics such as Equal Error Rate (EER %), we incorporated five established fairness metrics to quantify gender disparities in the model. Our results show that even when the overall EER difference between genders appears low, fairness-aware evaluation reveals disparities in error distribution that are obscured by aggregate performance measures. These findings demonstrate that reliance on standard metrics is unreliable, whereas fairness metrics provide critical insights into demographic-specific failure modes. This work highlights the importance of fairness-aware evaluation for developing a more equitable, robust, and trustworthy audio deepfake detection system.