Pushing the Performance of Synthetic Speech Detection with Kolmogorov-Arnold Networks and Self-Supervised Learning Models

2025-06-17

AI Summary

To address the growing threat of highly realistic synthetic speech in spoofing attacks, this work integrates Kolmogorov–Arnold Networks (KANs) into the XLSR-Conformer architecture, replacing conventional MLP layers to enhance discriminative capability in self-supervised speech representations. This is the first application of KANs to synthetic speech detection, leveraging their theoretically grounded nonlinear function approximation capacity to improve deep feature modeling in SSL models. Evaluated on the ASVspoof2021 benchmark, the proposed method achieves a 60.55% relative reduction in equal error rate (EER) on both the Logical Access (LA) and Deepfake (DF) subsets, with an EER of 0.70% on LA—substantially outperforming existing self-supervised baselines. These results demonstrate the effectiveness and generalization potential of KANs for audio deepfake detection.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Unsupervised & Self-Supervised Learning
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

Recent advancements in speech synthesis technologies have led to increasingly advanced spoofing attacks, posing significant challenges for automatic speaker verification systems. While systems based on self-supervised learning (SSL) models, particularly the XLSR-Conformer model, have demonstrated remarkable performance in synthetic speech detection, there remains room for architectural improvements. In this paper, we propose a novel approach that replaces the traditional Multi-Layer Perceptron in the XLSR-Conformer model with a Kolmogorov-Arnold Network (KAN), a novel architecture based on the Kolmogorov-Arnold representation theorem. Our results on ASVspoof2021 demonstrate that integrating KAN into the SSL-based models can improve the performance by 60.55% relatively on LA and DF sets, further achieving 0.70% EER on the 21LA set. These findings suggest that incorporating KAN into SSL-based models is a promising direction for advances in synthetic speech detection.

Problem

• Detect synthetic speech to counter spoofing attacks
• Improve SSL-based models for better detection performance
• Enhance architecture using Kolmogorov-Arnold Networks (KAN)

Innovation

• Replacing MLP with Kolmogorov-Arnold Network
• Integrating KAN into SSL-based models
• Improving synthetic speech detection performance