SZU-AFS Antispoofing System For The Asvspoof 5 Challenge

Abstract

This paper presents the SZU-AFS anti-spoofing system, designed for Track 1 of the ASVspoof 5 Challenge under open conditions. The system is built with four stages: selecting a baseline model, exploring effective data augmentation (DA) methods for fine-tuning, applying a co-enhancement strategy based on gradient norm aware minimization (GAM) for secondary fine-tuning, and fusing logits scores from the two best-performing fine-tuned models. The system utilizes the Wav2Vec2 front-end feature extractor and the AASIST back-end classifier as the baseline model. During model fine-tuning, three distinct DA policies have been investigated: single-DA, random-DA, and cascade-DA. Moreover, the employed GAM-based co-enhancement strategy, designed to fine-tune the augmented model at both data and optimizer levels, helps the Adam optimizer find flatter minima, thereby boosting model generalization. Overall, the final fusion system achieves a minDCF of 0.115 and an EER of 4.04% on the evaluation set.

Stats

• citations4
• S2 citations—
• github stars0
• HF likes0
• heat score5.24
• arxiv keyxu2024szu

Related papers

• The DKU Replay Detection System For The Asvspoof 2019 Challenge: On Data Augmentation, Feature Representation, Classification, And Fusion (2019)12.25
• Asasvicomtech: The Vicomtech-ugr Speech Deepfake Detection And SASV Systems For The Asvspoof5 Challenge (2024)5.24
• Spoofing-robust Speaker Verification Using Parallel Embedding Fusion: BTU Speech Group's Approach For Asvspoof5 Challenge (2024)5.24
• Temporal Variability And Multi-viewed Self-supervised Representations To Tackle The Asvspoof5 Deepfake Challenge (2024)0.00
• SA-SASV: An End-to-end Spoof-aggregated Spoofing-aware Speaker Verification System (2022)6.77
• Joint Optimization Of Speaker And Spoof Detectors For Spoofing-robust Automatic Speaker Verification (2025)0.00
• Asvspoof 5: Design, Collection And Validation Of Resources For Spoofing, Deepfake, And Adversarial Attack Detection Using Crowdsourced Speech (2025)9.76
• ASSERT: Anti-spoofing With Squeeze-excitation And Residual Networks (2019)15.40