WaveSP-Net: Learnable Wavelet-Domain Sparse Prompt Tuning for Speech Deepfake Detection

2025-10-06

AI Summary

Current front-end approaches for deepfake speech detection rely on full fine-tuning of large pre-trained models (e.g., XLSR), suffering from low parameter efficiency and poor generalization. To address this, we propose a parameter-efficient learnable wavelet-domain sparse prompt tuning method: the first to integrate wavelet transforms with prompt tuning, injecting prompt embeddings into multi-resolution features to precisely localize synthetic artifacts while keeping the pre-trained model frozen. We design a Partial-WSPT-XLSR front-end coupled with a bidirectional Mamba back-end, enabling multi-scale feature extraction and highly efficient adaptation. Our method trains only a minimal number of parameters—less than 0.1% of the base model—and achieves state-of-the-art performance on the newly released Deepfake-Eval-2024 and SpoofCeleb benchmarks. Crucially, it demonstrates significantly improved generalization to real-world scenarios, outperforming prior methods by substantial margins in both accuracy and robustness.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Speech
• Computer Vision: Diffusion Models for Vision

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Modern front-end design for speech deepfake detection relies on full fine-tuning of large pre-trained models like XLSR. However, this approach is not parameter-efficient and may lead to suboptimal generalization to realistic, in-the-wild data types. To address these limitations, we introduce a new family of parameter-efficient front-ends that fuse prompt-tuning with classical signal processing transforms. These include FourierPT-XLSR, which uses the Fourier Transform, and two variants based on the Wavelet Transform: WSPT-XLSR and Partial-WSPT-XLSR. We further propose WaveSP-Net, a novel architecture combining a Partial-WSPT-XLSR front-end and a bidirectional Mamba-based back-end. This design injects multi-resolution features into the prompt embeddings, which enhances the localization of subtle synthetic artifacts without altering the frozen XLSR parameters. Experimental results demonstrate that WaveSP-Net outperforms several state-of-the-art models on two new and challenging benchmarks, Deepfake-Eval-2024 and SpoofCeleb, with low trainable parameters and notable performance gains. The code and models are available at https://github.com/xxuan-acoustics/WaveSP-Net.

Problem

• Parameter-efficient speech deepfake detection front-end design
• Enhancing generalization to realistic in-the-wild audio data
• Localizing subtle synthetic artifacts without altering frozen parameters

Innovation

• Parameter-efficient front-ends fuse prompt-tuning with signal processing
• WaveSP-Net combines wavelet-domain sparse prompt tuning and Mamba back-end
• Multi-resolution features enhance artifact localization without altering frozen parameters