TY - GEN
T1 - Variational Regularization for End-to-End Speech Deepfake Detection
AU - Qin, Siqing
AU - Lee, Kong Aik
AU - Mak, Man Wai
AU - Lisena, Pasquale
AU - Todisco, Massimiliano
N1 - Publisher Copyright:
© 2025 IEEE.
PY - 2025/10
Y1 - 2025/10
N2 - Current research in end-to-end speech deepfake detection predominantly centers around inputting 'raw' waveforms to a deep architecture, such as RawNet2, and training the deep neural network to predict if the waveforms are fake. However, direct processing of waveforms could cause over-parameterization in the network, reducing its generalizability. To overcome this limitation, we propose a multi-level variational regularization framework integrating a modified Variational Autoencoder (VAE) with discriminative constraints. Specifically, we adopt an VAE with a deepfake discrimination constraint to regularize a RawNet2-based high-level feature map (HFM) extractor. Experimental results show that the proposed variational regularization leads to HFM features that improve the performance of AASIST, SE-Rawformer, and RawBMamba by 3 6. 0 1%, } {1 0.07%, and 6.35%, respectively.
AB - Current research in end-to-end speech deepfake detection predominantly centers around inputting 'raw' waveforms to a deep architecture, such as RawNet2, and training the deep neural network to predict if the waveforms are fake. However, direct processing of waveforms could cause over-parameterization in the network, reducing its generalizability. To overcome this limitation, we propose a multi-level variational regularization framework integrating a modified Variational Autoencoder (VAE) with discriminative constraints. Specifically, we adopt an VAE with a deepfake discrimination constraint to regularize a RawNet2-based high-level feature map (HFM) extractor. Experimental results show that the proposed variational regularization leads to HFM features that improve the performance of AASIST, SE-Rawformer, and RawBMamba by 3 6. 0 1%, } {1 0.07%, and 6.35%, respectively.
UR - https://www.scopus.com/pages/publications/105030493850
U2 - 10.1109/APSIPAASC65261.2025.11249209
DO - 10.1109/APSIPAASC65261.2025.11249209
M3 - Conference article published in proceeding or book
AN - SCOPUS:105030493850
T3 - 2025 Asia Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2025
SP - 2241
EP - 2246
BT - 2025 Asia Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2025
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 17th Asia Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA ASC 2025
Y2 - 22 October 2025 through 24 October 2025
ER -