Generalizable Speech Deepfake Detection via Information Bottleneck Enhanced Adversarial Alignment

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.

Key Contributions

• Formalizes speech deepfake detection as a dual distribution shift problem (covariate shift + concept shift) and proposes attack-invariant feature learning as the solution
• Proposes IB-CAAN: confidence-guided adversarial alignment that selectively suppresses attack-specific artifacts while preserving discriminative cues, combined with an information bottleneck to compress nuisance variability
• Achieves state-of-the-art performance on ASVspoof 2019/2021, ASVspoof 5, and In-the-Wild benchmarks, demonstrating improved generalization to unseen spoofing methods

🛡️ Threat Analysis

Details

Similar Papers