AdvSV
收藏arXiv2024-01-16 更新2024-08-06 收录
下载链接：
http://arxiv.org/abs/2310.05369v2
下载链接
链接失效反馈官方服务：
资源简介：
AdvSV数据集是为研究自动语音验证（ASV）系统对抗性攻击而创建的开放源代码数据集。该数据集基于Voxceleb1验证测试集构建，包含314,496个样本，旨在模拟空中攻击设置。数据集创建过程中使用了代表性的ASV模型，并通过扰动生成算法、扬声器、麦克风和声学环境来生成对抗样本。AdvSV数据集主要应用于解决ASV系统在对抗性攻击下的安全问题，特别是在可重复性研究中提供标准数据集。
The AdvSV dataset is an open-source dataset developed for research on adversarial attacks against automatic speaker verification (ASV) systems. Built upon the VoxCeleb1 verification test set, it contains 314,496 samples and is designed to simulate over-the-air adversarial attack scenarios. During the dataset construction, representative ASV models, perturbation generation algorithms, real-world speakers, microphones, and acoustic environments were utilized to generate adversarial samples. The AdvSV dataset is mainly used to address the security issues of ASV systems under adversarial attacks, and specifically serves as a standard benchmark dataset for reproducible research.
提供机构：
数据科学与大数据技术学院，深圳大数据研究院，香港中文大学（深圳）
创建时间：
2023-10-09
