Domain-Adaptive Dual-Gating Mixture of Experts for Generalizable Speech Deepfake Detection

2026-09-27

AI Summary

This study addresses the limited generalization of gating networks to unseen attacks in audio deepfake detection, caused by their neglect of acoustic temporal cues. To overcome this, we propose a domain-adaptive dual-gated Mixture-of-Experts (MoE) framework. Methodologically, the approach fuses raw waveforms with self-supervised learning (SSL) features and employs Sinc-layer filters to extract multi-granularity signals. Furthermore, it introduces an innovative dual-gating mechanism incorporating domain prototypes to enable intelligent expert routing based on implicit forgery patterns. Experimental results demonstrate that the proposed model significantly outperforms baselines on challenging datasets, achieving a 40.8% relative reduction in equal error rate. These findings highlight its superior cross-domain generalization capability and architectural efficiency.

Abstract

Recent advances in speech deepfake detection (SDD) have leveraged the Mixture of Experts (MoE) to enhance generalization capacity. However, existing gating networks often overlook the acoustic and temporal cues of deepfakes. In this work, we propose a novel domain-adaptive dual-gating MoE (DADGMoE) framework for SDD under unseen attack types and acoustic conditions. Our innovative dual-gating mechanism leverages Sinc-layer-based filters to process both low-level acoustic signals (raw waveforms) and high-level speech representations from a large self-supervised learning (SSL) model. It further incorporates domain prototypes to guide expert routing based on implicit deepfake patterns. The lightweight affine experts process the routed inputs. Experiments show that our DADGMoE significantly outperforms the baseline, achieving up to a 40.8% relative EER reduction on challenging out-of-dataset benchmarks. This framework demonstrates superior generalization capabilities and efficient design.

Problem

• Speech Deepfake Detection
• Domain Adaptation
• Generalization
• Mixture of Experts

Innovation

• Mixture of Experts
• Dual-Gating Mechanism
• Domain Adaptation
• Speech Deepfake Detection
• Self-Supervised Learning