Spoofing-Aware Speaker Verification via Wavelet Prompt Tuning and Multi-Model Ensembles

2026-01-24

AI Summary

This work addresses the threat posed by generative speech spoofing to speaker verification systems by proposing an anti-spoofing-aware joint discrimination framework. The approach integrates a wavelet prompt-tuned XLSR-AASIST anti-spoofing module with a multi-model ensemble speaker verification system comprising ResNet34, ResNet293, and WavLM-ECAPA-TDNN, optimized end-to-end via Z-score normalization and score averaging. A novel wavelet prompt tuning mechanism is introduced to significantly enhance robustness against spoofed utterances. Evaluated in the WildSpoof 2026 Challenge, the system achieves a Macro a-DCF of 0.2017, an SASV EER of 2.08%, and an impressively low within-domain spoof detection EER of 0.16%, demonstrating its effectiveness while highlighting cross-domain generalization as a persistent challenge.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Adversarial Agents

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems

Abstract

This paper describes the UZH-CL system submitted to the SASV section of the WildSpoof 2026 challenge. The challenge focuses on the integrated defense against generative spoofing attacks by requiring the simultaneous verification of speaker identity and audio authenticity. We proposed a cascaded Spoofing-Aware Speaker Verification framework that integrates a Wavelet Prompt-Tuned XLSR-AASIST countermeasure with a multi-model ensemble. The ASV component utilizes the ResNet34, ResNet293, and WavLM-ECAPA-TDNN architectures, with Z-score normalization followed by score averaging. Trained on VoxCeleb2 and SpoofCeleb, the system obtained a Macro a-DCF of 0.2017 and a SASV EER of 2.08%. While the system achieved a 0.16% EER in spoof detection on the in-domain data, results on unseen datasets, such as the ASVspoof5, highlight the critical challenge of cross-domain generalization.

Problem

• speaker verification
• spoofing detection
• cross-domain generalization
• audio authenticity
• generative spoofing attacks

Innovation

• Wavelet Prompt Tuning
• Multi-Model Ensemble
• Spoofing-Aware Speaker Verification
• Cross-Domain Generalization
• XLSR-AASIST