你所训练的就是你所得到的：音频深度伪造检测中的性别偏见、训练构成及事后缓解
What You Train Is What You Get: Gender Bias, Training Composition, and Post-Hoc Mitigation in Audio Deepfake Detection
浏览论文内容
中文总结 AI 辅助
研究音频深度伪造检测中的性别偏见，通过在不同性别构成训练集上训练特定攻击模型，使用ResNet18分类器及评估六种校准方法，发现训练数据构成影响偏差方向，事后方法只能部分缓解差异，强调性别公平需在训练时解决。
中文摘要 AI 辅助
音频深度伪造检测模型用于判断语音是真实的还是人工生成的，但高总体准确率可能掩盖不同人口群体间的显著性能差异。在这项工作中，我们使用ASVspoof5数据集研究音频深度伪造检测中的性别偏见。我们在一个旨在分离性别构成影响的可控自定义划分下使用ASVspoof5。我们在九个不同性别构成的训练集上训练特定攻击模型，从全女性到全男性。我们使用带有对数频谱图和WavLM - Base +特征的ResNet18分类器，并评估六种事后阈值校准方法。实验结果表明，训练数据构成强烈预测偏差方向，代表性不足的性别在测试时表现更差。在相同训练条件下，WavLM - Base +特征产生的性别性能差距比对数频谱图大3.0到4.3倍，平衡训练可减少对数频谱图偏差，但WavLM偏差基本不变。此外，所有六种校准策略，包括可访问完整测试集标签的Oracle校准，等错误率差距在1.317个百分点保持不变，这证实阈值调整无法纠正潜在的分数分布差异。总体而言，这些发现表明音频深度伪造检测中的性别公平必须在训练时解决，因为事后方法只能部分缓解由此产生的差异。
英文摘要
Audio deepfake detection models determine whether speech is genuine or artificially generated, but high overall accuracy can mask substantial performance disparities across demographic groups. In this work, we investigate gender bias in audio deepfake detection using the ASVspoof5 dataset. We use ASVspoof5 under a controlled custom split designed to isolate gender-composition effects. We train attack-specific models on nine training sets with different gender compositions, ranging from female-only to male-only. We use a ResNet18 classifier with LogSpectrogram and WavLM-Base+ features, and we evaluated six post-hoc threshold calibration methods. Experimental results show that training data composition strongly predicts bias direction, with the underrepresented gender performing worse at test time. WavLM-Base+ features are shown to produce gender performance gaps 3.0 to 4.3 times larger than LogSpectrogram under identical training conditions, and balanced training is found to reduce LogSpectrogram bias but leave WavLM bias largely intact. Moreover, all six calibration strategies, including Oracle calibration with full test-set label access, leave the Equal Error Rate gap unchanged at 1.317 pp, confirming that threshold adjustment cannot correct underlying score distribution disparities. Overall, these findings suggest that gender fairness in audio deepfake detection must be addressed at training time, as post-hoc methods can only partially mitigate the resulting disparities
发表机构
- School of Computing, Wichita State University（威奇托州立大学计算学院）
- Institut national de la recherche scientifique (INRS–EMT)（国家科学研究院（INRS–EMT））
- INRS-UQO Mixed Research Unit on Cybersecurity（INRS-UQO网络安全混合研究单位）
机构由 AI 辅助整理，请以论文原文为准。
