Amplifying Artifacts with Speech Enhancement in Voice Anti-spoofing

2025-06-13

AI Summary

To address the challenge of detecting subtle artifacts in generative voice spoofing, this paper proposes a model-agnostic, three-stage active amplification pipeline: (1) injecting controlled noise into input speech; (2) leveraging off-the-shelf speech enhancement models (e.g., DCCRN, SEGAN) to disentangle and isolate noise from spoofing artifacts; and (3) selectively amplifying spoofing traces in the feature domain. This work pioneers the repurposing of speech enhancement—from passive noise suppression to active artifact enhancement—without modifying downstream detector architectures and maintaining compatibility with diverse anti-spoofing frameworks. Evaluated on ASVspoof2019 and ASVspoof2021, the method achieves up to 44.44% and 26.34% relative reduction in Equal Error Rate (EER), respectively, and significantly improves cross-generation-model generalization performance.

Technology Category

• Machine Learning: Active Learning
• Natural Language Processing: Generation
• Computer Vision: Generative Adversarial Networks (GANs) for Vision

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Privacy-enhancing technologies
• Search and Retrieval-Augmented AI: Agentic search

Abstract

Spoofed utterances always contain artifacts introduced by generative models. While several countermeasures have been proposed to detect spoofed utterances, most primarily focus on architectural improvements. In this work, we investigate how artifacts remain hidden in spoofed speech and how to enhance their presence. We propose a model-agnostic pipeline that amplifies artifacts using speech enhancement and various types of noise. Our approach consists of three key steps: noise addition, noise extraction, and noise amplification. First, we introduce noise into the raw speech. Then, we apply speech enhancement to extract the entangled noise and artifacts. Finally, we amplify these extracted features. Moreover, our pipeline is compatible with different speech enhancement models and countermeasure architectures. Our method improves spoof detection performance by up to 44.44% on ASVspoof2019 and 26.34% on ASVspoof2021.

Problem

• Detect hidden artifacts in spoofed speech
• Amplify artifacts using speech enhancement
• Improve spoof detection performance significantly

Innovation

• Amplify artifacts using speech enhancement
• Model-agnostic pipeline with noise addition
• Compatible with various enhancement models