Do Compact SSL Backbones Matter for Audio Deepfake Detection? A Controlled Study with RAPTOR
This research is significant for developers and researchers in audio deepfake detection, as it identifies the importance of SSL pre-training trajectory over model scale for reliable detection, potentially enabling more efficient and robust systems.
This study investigates the impact of compact self-supervised learning (SSL) backbones on audio deepfake detection, using a unified pairwise-gated fusion detector called RAPTOR. It found that multilingual HuBERT pre-training is crucial for cross-domain robustness, allowing 100M models to perform comparably to larger and commercial systems.
Self-supervised learning (SSL) underpins modern audio deepfake detection, yet most prior work centers on a single large wav2vec2-XLSR backbone, leaving compact under studied. We present RAPTOR, Representation Aware Pairwise-gated Transformer for Out-of-domain Recognition a controlled study of compact SSL backbones from the HuBERT and WavLM within a unified pairwise-gated fusion detector, evaluated across 14 cross-domain benchmarks. We show that multilingual HuBERT pre-training is the primary driver of cross-domain robustness, enabling 100M models to match larger and commercial systems. Beyond EER, we introduce a test-time augmentation protocol with perturbation-based aleatoric uncertainty to expose calibration differences invisible to standard metrics: WavLM variants exhibit overconfident miscalibration under perturbation, whereas iterative mHuBERT remains stable. These findings indicate that SSL pre-training trajectory, not model scale, drives reliable audio deepfake detection.