Source-Directed Trajectory Perturbation at First-Order Cost for Domain Generalization in Speech Deepfake Detection

Authors: Siqing Qin, Kong Aik Lee, Youzhi Tu

Published: 2026-10-07 13:51:31+00:00

AI Summary

This paper addresses the challenge of domain generalization in speech deepfake detection by introducing two worst-case meta-learning for domain generalization (WC-MLDG) variants. WC-MLDG-4P explicitly perturbs meta-train and meta-test states, while the more efficient WC-MLDG-2P uses a source-directed trajectory perturbation strategy. WC-MLDG-2P achieves significant EER reductions compared to standard MLDG, offering a robust solution at a lower computational cost.

Abstract

Speech deepfake detectors often lose accuracy when the distribution of the test data differs from that of the training data. Meta-learning for domain generalization (MLDG) shows promise by simulating domain shifts with episodic meta-train and meta-test splits. However, the MLDG meta-objective considers only a single clean adaptation trajectory and does not account for the local loss landscape around its endpoints. To address this limitation, we first propose an explicit worst-case MLDG variant, dubbed WC-MLDG-4P, which perturbs both the meta-train and meta-test states but requires four gradient evaluations per episode. We then introduce WC-MLDG-2P, a source-directed alternative to the explicit robust objective. It shifts the clean MLDG endpoint toward a locally higher source-loss state and evaluates the meta-test gradient there, retaining the two-evaluation cost of the first-order MLDG. A first-order expansion relates the resulting gradient change to meta-test directional curvature along the source gradient without explicitly computing a Hessian. Relative to MLDG, WC-MLDG-2P achieves relative mean-EER reductions of 23.4% with XLSR-AASIST and 7.6% with XLSR-Conformer-TCM, respectively.