Assessing the Impact of Speaker Identity in Speech Spoofing Detection

Spoofing detection systems are typically trained using diverse recordings from multiple speakers, often assuming that the resulting embeddings are independent of speaker identity. However, this assumption remains unverified. In this paper, we investigate the impact of speaker information on spoofing detection systems. We propose two approaches within our Speaker-Invariant Multi-Task framework, one that models speaker identity within the embeddings and another that removes it. SInMT integrates multi-task learning for joint speaker recognition and spoofing detection, incorporating a gradient reversal layer. Evaluated using four datasets, our speaker-invariant model reduces the average equal error rate by 17% compared to the baseline, with up to 48% reduction for the most challenging attacks (e.g., A11).

Key Contributions

• SInMT framework that unifies speaker-aware and speaker-invariant spoofing detection via a dual-head MHFA classifier with optional gradient reversal layer (GRL)
• Empirical demonstration that removing speaker identity information (via GRL) reduces average EER by 17.2% across four datasets and 48% for the hardest attack type (A11)
• Analysis of how explicit speaker identity modeling (speaker-aware) vs. suppression (speaker-invariant) each improve over the baseline MHFA spoofing detector

🛡️ Threat Analysis

Details

Similar Papers