LlamaPartialSpoof：一个由 LLM 驱动、模拟虚假信息生成的假语音数据集
LlamaPartialSpoof: An LLM-Driven Fake Speech Dataset Simulating Disinformation Generation
AI总结：
本文构建 130 小时的 LlamaPartialSpoof 数据集，利用 LLM 与语音克隆模拟全假及部分假语音，揭示当前假语音检测 CM 在未知场景下的泛化漏洞。
摘要：
以往的假语音数据集是从防御者视角构建的，用于开发对策（CM）系统，而没有考虑攻击者的多样动机。为了更好地贴合现实场景，我们创建了 LlamaPartialSpoof，这是一个包含全假语音和部分假语音的 130 小时数据集，使用大型语言模型（LLM）和语音克隆技术来评估 CM 的鲁棒性。通过考察对攻击者和防御者都有价值的信息，我们发现了当前 CM 系统中的若干关键漏洞，这些漏洞可被利用以提高攻击成功率，包括对某些文本转语音模型或拼接方法的偏向。实验结果表明，当前假语音检测系统难以泛化到未见过的场景，取得的最佳性能为 24.49% 等错误率。
英文摘要：
Previous fake speech datasets were constructed from a defender's perspective to develop countermeasure (CM) systems without considering diverse motivations of attackers. To better align with real-life scenarios, we created LlamaPartialSpoof, a 130-hour dataset that contains both fully and partially fake speech, using a large language model (LLM) and voice cloning technologies to evaluate the robustness of CMs. By examining valuable information for both attackers and defenders, we identify several key vulnerabilities in current CM systems, which can be exploited to enhance attack success rates, including biases toward certain text-to-speech models or concatenation methods. Our experimental results indicate that the current fake speech detection system struggle to generalize to unseen scenarios, achieving a best performance of 24.49% equal error rate.
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。
