E-branchformer improves audio deepfake detection with local and global features
Disentangled Global-Local Feature Learning with E-Branchformer for Audio Deepfake Detection
Sound
Summary
Voice cloning technology can trick systems that recognize real human voices. This paper presents a new tool called E-Branchformer that listens to audio and checks if it is fake or not. It looks at the sound in two ways at the same time—catching the overall picture and the small details. Tests show this method works better than previous ones on different collections of voice recordings. The design helps it detect fake voices even in new, tricky situations.
What this means in practice
- •For security teams: Improve voice authentication security by accurately detecting fake audio attacks in real-world environments.
- •For speech technology developers: Build more reliable voice synthesis and verification pipelines by integrating robust deepfake detection using dual-branch models.
Authors
Phuong Tuan Dat, Ho Bao Thu, Nguyen Tran Trung, Pham Viet Hoang, Nguyen Thi Thu Trang
Abstract
The rapid advancement of voice synthesis technologies such as text-to-speech and voice conversion poses significant threats to speech-based authentication systems, necessitating robust deepfake detection methods. In this work, we propose a novel E-Branchformer-based architecture that effectively leverages self-supervised speech representations for audio deepfake detection. Our model employs parallel branches to simultaneously capture global contextual dependencies through multi-head self-attention and local temporal patterns through convolutional processing. To enhance discriminative capability, we integrate depthwise convolution and Squeeze-and-Excitation modules that enrich the classification token with refined patch token information after feature merging. Extensive experiments on ASVspoof 2021 LA, DF, and In-the-Wild datasets demonstrate state-of-the-art performance with equal error rates of 0.88%, 1.85%, and 6.30% respectively, substantially outperforming existing methods. Comprehensive ablation studies validate that the dual-branch architecture provides complementary discriminative information, Squeeze-and-Excitation Aggregation significantly improves SSL feature integration, and the combination of DWConv and SE modules is critical for effective class token enhancement. The superior performance on real-world scenarios demonstrates strong generalization capability to diverse acoustic conditions and unseen spoofing attacks.