Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2509.23618 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866915519718752256 |
|---|---|
| author | Huang, Pu Wang, Shouguang Yao, Siya Zhou, Mengchu |
| author_facet | Huang, Pu Wang, Shouguang Yao, Siya Zhou, Mengchu |
| contents | Neural speech synthesis techniques have enabled highly realistic speech deepfakes, posing major security risks. Speech deepfake detection is challenging due to distribution shifts across spoofing methods and variability in speakers, channels, and recording conditions. We explore learning shared discriminative features as a path to robust detection and propose Information Bottleneck enhanced Confidence-Aware Adversarial Network (IB-CAAN). Confidence-guided adversarial alignment adaptively suppresses attack-specific artifacts without erasing discriminative cues, while the information bottleneck removes nuisance variability to preserve transferable features. Experiments on ASVspoof 2019/2021, ASVspoof 5, and In-the-Wild demonstrate that IB-CAAN consistently outperforms baseline and achieves state-of-the-art performance on many benchmarks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_23618 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Generalizable Speech Deepfake Detection via Information Bottleneck Enhanced Adversarial Alignment Huang, Pu Wang, Shouguang Yao, Siya Zhou, Mengchu Sound Artificial Intelligence Neural speech synthesis techniques have enabled highly realistic speech deepfakes, posing major security risks. Speech deepfake detection is challenging due to distribution shifts across spoofing methods and variability in speakers, channels, and recording conditions. We explore learning shared discriminative features as a path to robust detection and propose Information Bottleneck enhanced Confidence-Aware Adversarial Network (IB-CAAN). Confidence-guided adversarial alignment adaptively suppresses attack-specific artifacts without erasing discriminative cues, while the information bottleneck removes nuisance variability to preserve transferable features. Experiments on ASVspoof 2019/2021, ASVspoof 5, and In-the-Wild demonstrate that IB-CAAN consistently outperforms baseline and achieves state-of-the-art performance on many benchmarks. |
| title | Generalizable Speech Deepfake Detection via Information Bottleneck Enhanced Adversarial Alignment |
| topic | Sound Artificial Intelligence |
| url | https://arxiv.org/abs/2509.23618 |