一种基于分层中心引导加权聚合的联邦深度伪造语音检测方法
A Federated Deepfake Speech Detection Method Based on Layer-Wise Center-Guided Weighting Aggregation
浏览论文内容
中文总结 AI 辅助
提出联邦深度伪造语音检测方法FedDSD，采用FedProx训练与分层中心引导加权聚合策略，在保护隐私的同时实现与集中式训练相当的检测性能，并展现跨域泛化能力。
中文摘要 AI 辅助
基于深度学习的语音合成技术的进步显著增加了深度伪造语音的多样性，对语音认证构成了威胁。虽然集中式训练对于深度伪造语音检测（DSD）是有效的，但它需要大量的计算资源并引发隐私问题。为了解决这些问题，我们提出了一种联邦深度伪造语音检测（FedDSD）方法，该方法能够在无需共享原始音频的情况下，跨分散的语音数据集进行协作模型训练。具体而言，每个客户端使用FedProx算法训练本地模型，以减轻数据异质性的影响，并将模型参数上传到中央服务器。为了改进全局模型聚合，我们进一步提出了一种分层中心引导加权聚合（L-CGWA）策略，该策略根据每个客户端到参考中心的距离，逐层调整其对全局模型的贡献，捕捉客户端间和层间的差异，从而增强模型聚合的鲁棒性。实验结果表明，在提出的FedDSD方法下训练的模型实现了与集中式协同训练相当的平均错误率（EERs），同时显著优于在单个语料库上训练的模型。此外，所提出的FedDSD方法在多样的跨领域数据集上展示了强大的泛化能力。
英文摘要
The advancement of deep learning-based speech synthesis has significantly increased the diversity of deepfake speech, posing threats to voice authentication. While centralized training is effective for deepfake speech detection (DSD), it requires considerable computational resources and raises privacy concerns. To address these issues, we propose a Federated DSD (FedDSD) method that enables collaborative model training across decentralized speech datasets without sharing raw audio. Specifically, each client trains a local model using the FedProx algorithm to mitigate the effects of data heterogeneity and uploads model parameters to a central server. To improve global model aggregation, we further propose a layer-wise center-guided weighting aggregation (L-CGWA) strategy that adjusts each client's contribution per layer based on its distance to a reference center, capturing inter-client and inter-layer discrepancies and enhancing the robustness of model aggregation. Experimental results demonstrate that models trained under the proposed FedDSD method achieve equal error rates (EERs) comparable to those obtained via centralized co-training, while significantly out-performing models trained on individual corpora. Furthermore, the proposed FedDSD method demonstrates robust generalization capabilities across diverse cross-domain datasets.
发表机构
- School of Communications and Information Engineering, Nanjing University of Posts and Telecommunications（南京邮电大学通信与信息工程学院）
机构由 AI 辅助整理，请以论文原文为准。
