arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2406.04582eess.AScs.SD
收藏

基于神经编解码器的说话人验证对抗样本检测

Neural Codec-based Adversarial Sample Detection for Speaker Verification

National Taiwan University(台湾大学)

Xuanjun Chen, Jiawei Du, Haibin Wu, Jyh-Shing Roger Jang, Hung-yi Lee

AI总结:

本文提出基于神经编解码器的说话人验证对抗样本检测方法,通过比较原始与重合成音频的ASV得分差异进行检测,其中Descript-audio-codec在15个编解码器中检测率最高,并超越多种SOTA方法。

摘要:

自动说话人验证(ASV)在安全关键应用中日益普及,但面临日益增多的对抗攻击,目前有效的防御手段较少。本文提出一种基于神经编解码器的ASV对抗样本检测方法。该方法利用编解码器丢弃冗余扰动并保留关键信息的能力。具体而言,我们通过比较原始音频与经编解码器模型重新合成音频之间的ASV得分差异来区分真实样本和对抗样本。本研究全面探索了所有开源神经编解码器及其变体模型进行实验。其中,Descript-audio-codec模型表现突出,在15个神经编解码器中取得了最高检测率,并超越了七种先前最先进(SOTA)的检测方法。值得注意的是,我们的单模型方法甚至大幅优于一种SOTA集成方法。

英文摘要:

Automatic Speaker Verification (ASV), increasingly used in security-critical applications, faces vulnerabilities from rising adversarial attacks, with few effective defenses available. In this paper, we propose a neural codec-based adversarial sample detection method for ASV. The approach leverages the codec's ability to discard redundant perturbations and retain essential information. Specifically, we distinguish between genuine and adversarial samples by comparing ASV score differences between original and re-synthesized audio (by codec models). This comprehensive study explores all open-source neural codecs and their variant models for experiments. The Descript-audio-codec model stands out by delivering the highest detection rate among 15 neural codecs and surpassing seven prior state-of-the-art (SOTA) detection methods. Note that, our single-model method even outperforms a SOTA ensemble method by a large margin.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑