BiCrossMamba-ST: Speech Deepfake Detection with Bidirectional Mamba Spectro-Temporal Cross-Attention

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kheir, Yassine El, Polzehl, Tim, Möller, Sebastian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913848115593216
author Kheir, Yassine El
Polzehl, Tim
Möller, Sebastian
author_facet Kheir, Yassine El
Polzehl, Tim
Möller, Sebastian
contents We propose BiCrossMamba-ST, a robust framework for speech deepfake detection that leverages a dual-branch spectro-temporal architecture powered by bidirectional Mamba blocks and mutual cross-attention. By processing spectral sub-bands and temporal intervals separately and then integrating their representations, BiCrossMamba-ST effectively captures the subtle cues of synthetic speech. In addition, our proposed framework leverages a convolution-based 2D attention map to focus on specific spectro-temporal regions, enabling robust deepfake detection. Operating directly on raw features, BiCrossMamba-ST achieves significant performance improvements, a 67.74% and 26.3% relative gain over state-of-the-art AASIST on ASVSpoof LA21 and ASVSpoof DF21 benchmarks, respectively, and a 6.80% improvement over RawBMamba on ASVSpoof DF21. Code and models will be made publicly available.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13930
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle BiCrossMamba-ST: Speech Deepfake Detection with Bidirectional Mamba Spectro-Temporal Cross-Attention
Kheir, Yassine El
Polzehl, Tim
Möller, Sebastian
Sound
Audio and Speech Processing
We propose BiCrossMamba-ST, a robust framework for speech deepfake detection that leverages a dual-branch spectro-temporal architecture powered by bidirectional Mamba blocks and mutual cross-attention. By processing spectral sub-bands and temporal intervals separately and then integrating their representations, BiCrossMamba-ST effectively captures the subtle cues of synthetic speech. In addition, our proposed framework leverages a convolution-based 2D attention map to focus on specific spectro-temporal regions, enabling robust deepfake detection. Operating directly on raw features, BiCrossMamba-ST achieves significant performance improvements, a 67.74% and 26.3% relative gain over state-of-the-art AASIST on ASVSpoof LA21 and ASVSpoof DF21 benchmarks, respectively, and a 6.80% improvement over RawBMamba on ASVSpoof DF21. Code and models will be made publicly available.
title BiCrossMamba-ST: Speech Deepfake Detection with Bidirectional Mamba Spectro-Temporal Cross-Attention
topic Sound
Audio and Speech Processing
url https://arxiv.org/abs/2505.13930