Post-training for Deepfake Speech Detection

Publication
arXiv preprint arXiv:2506.21090 (Accepted at ASRU 2025)

This paper presents a study on improving deepfake speech detection via post-training strategies. The authors explore adaptation techniques that fine-tune detection models after initial training, focusing on robustness against unseen spoofing attacks and cross-domain generalization. The proposed methods demonstrate strong performance and were accepted at ASRU 2025.