SNAP: Speaker Nulling for Artifact Projection in Speech Deepfake Detection

TL;DR

SNAP is a novel framework that reduces speaker information in speech deepfake detection models, enabling better focus on artifacts and improving detection accuracy across unseen speakers.

Contribution

This paper introduces SNAP, a speaker-nulling method that isolates synthesis artifacts by removing speaker-dependent features, enhancing deepfake detection robustness.

Findings

SNAP achieves state-of-the-art detection performance.

Reducing speaker entanglement improves generalization to unseen speakers.

Orthogonal projection effectively suppresses speaker-specific information.

Abstract

Recent advancements in text-to-speech technologies enable generating high-fidelity synthetic speech nearly indistinguishable from real human voices. While recent studies show the efficacy of self-supervised learning-based speech encoders for deepfake detection, these models struggle to generalize across unseen speakers. Our quantitative analysis suggests these encoder representations are substantially influenced by speaker information, causing detectors to exploit speaker-specific correlations rather than artifact-related cues. We call this phenomenon speaker entanglement. To mitigate this reliance, we introduce SNAP, a speaker-nulling framework. We estimate a speaker subspace and apply orthogonal projection to suppress speaker-dependent components, isolating synthesis artifacts within the residual features. By reducing speaker entanglement, SNAP encourages detectors to focus on artifact-related patterns, leading to state-of-the-art performance.