arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新，休息一下吧，周末愉快~~
arXiv 2507.03468cs.SDeess.AS
收藏

部分伪造语音的鲁棒定位：指标与域外评估

Robust Localization of Partially Fake Speech: Metrics and Out-of-Domain Evaluation

Nanyang Technological University（南洋理工大学）；Ben Gurion University（本·古里安大学）；The Hong Kong Polytechnic University（香港理工大学）

Hieu-Thi Luong, Inbal Rimon, Haim Permuter, Kong Aik Lee, Eng Siong Chng

备注：
APSIPA 2025
阅读 PDF
arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文将部分伪造语音定位重构为序列异常检测，指出 EER 评估局限，并用 CFPRF 的域内外实验揭示模型泛化不足及部分伪造训练数据的重要性。

摘要：

部分音频深度伪造定位带来了独特挑战，并且与整句欺骗检测相比仍未得到充分研究。尽管近期方法报告了强劲的域内性能，但其在真实世界中的效用仍不明确。在本分析中，我们批判性地考察了当前评估实践的局限，尤其是 Equal Error Rate（EER，等错误率）的广泛使用；该指标往往会掩盖泛化能力和部署就绪程度。我们建议将定位任务重新表述为序列异常检测问题，并提倡使用依赖阈值的指标，如准确率、精确率、召回率和 F1 分数，因为它们能更好地反映真实世界行为。具体而言，我们分析了开源 Coarse-to-Fine Proposal Refinement Framework（CFPRF，由粗到细候选框优化框架）的性能：该模型在域内 PartialSpoof 评估集上取得 7.61% 的 20-ms EER，但在 LlamaPartialSpoof 和 Half-Truth 域外测试集上分别为 43.25% 和 27.59%。有趣的是，我们复现的同一模型版本在域内数据上表现更差（9.84%），但在域外集上表现更好（分别为 41.72% 和 14.98%）。这凸显了过度优化域内 EER 的风险，可能导致模型在真实场景中表现糟糕。这也表明，尽管深度学习模型在域内数据上可能有效，但它们向域外场景的泛化很差，无法检测新型合成样本，并会误分类不熟悉的真实音频。最后，我们观察到，向训练数据中加入更多真实或完全合成的话语往往会降低性能，而加入部分伪造话语则会提升性能。

英文摘要：

Partial audio deepfake localization poses unique challenges and remain underexplored compared to full-utterance spoofing detection. While recent methods report strong in-domain performance, their real-world utility remains unclear. In this analysis, we critically examine the limitations of current evaluation practices, particularly the widespread use of Equal Error Rate (EER), which often obscures generalization and deployment readiness. We propose reframing the localization task as a sequential anomaly detection problem and advocate for the use of threshold-dependent metrics such as accuracy, precision, recall, and F1-score, which better reflect real-world behavior. Specifically, we analyze the performance of the open-source Coarse-to-Fine Proposal Refinement Framework (CFPRF), which achieves a 20-ms EER of 7.61% on the in-domain PartialSpoof evaluation set, but 43.25% and 27.59% on the LlamaPartialSpoof and Half-Truth out-of-domain test sets. Interestingly, our reproduced version of the same model performs worse on in-domain data (9.84%) but better on the out-of-domain sets (41.72% and 14.98%, respectively). This highlights the risks of over-optimizing for in-domain EER, which can lead to models that perform poorly in real-world scenarios. It also suggests that while deep learning models can be effective on in-domain data, they generalize poorly to out-of-domain scenarios, failing to detect novel synthetic samples and misclassifying unfamiliar bona fide audio. Finally, we observe that adding more bona fide or fully synthetic utterances to the training data often degrades performance, whereas adding partially fake utterances improves it.

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。

↑