Explainable deepfake and spoofing detection: an attack analysis using SHapley Additive exPlanations

TL;DR

This paper uses SHAP explainability to analyze and identify artefacts in deepfake and spoofing detection classifiers, enhancing understanding of attack-specific features in synthetic speech and voice conversion attacks.

Contribution

It extends previous work by applying SHAP to analyze classifier behaviour and identify artefacts specific to different spoofing attack algorithms.

Findings

SHAP visualizations reveal attack-specific artefacts.

Differences between synthetic speech and voice conversion attacks are identified.

Classifiers detect artefacts in raw waveforms and spectrograms.

Abstract

Despite several years of research in deepfake and spoofing detection for automatic speaker verification, little is known about the artefacts that classifiers use to distinguish between bona fide and spoofed utterances. An understanding of these is crucial to the design of trustworthy, explainable solutions. In this paper we report an extension of our previous work to better understand classifier behaviour to the use of SHapley Additive exPlanations (SHAP) to attack analysis. Our goal is to identify the artefacts that characterise utterances generated by different attacks algorithms. Using a pair of classifiers which operate either upon raw waveforms or magnitude spectrograms, we show that visualisations of SHAP results can be used to identify attack-specific artefacts and the differences and consistencies between synthetic speech and converted voice spoofing attacks.