DeepFakeX

Title:DeepFakeX: A Comprehensive Dataset for Deepfake Research Description:DeepFakeX is a curated dataset comprising 800 videos across five categories: Face-swapped deepfakes (500) Audio-swapped deepfakes (100) Voice-cloned deepfakes (100) Combined face-and-audio-swapped deepfakes (100) The dataset was generated using a range of state-of-the-art deepfake creation tools to ensure high diversity and realism, while excluding political figures, minors, and other sensitive personal material to mitigate privacy and misuse risks. Each video is labeled with a type-specific prefix (ff_, fa_, vc_, ffa_) to indicate manipulation type. The dataset is provided as compressed archives for each category (e.g., deepfake_face_swapped_1.zip, deepfake_voice_cloned.zip). Specifically, a CSV metadata file has been included that provides structured information for each video sample, including video identifier, filename, deepfake category, generation method, language, contextual scenario, resolution, and duration. License:Creative Commons Attribution–NonCommercial 4.0 International (CC BY-NC 4.0) Disclaimer:This dataset is intended solely for non-commercial academic research on deepfake detection and multimedia forensics. Any use for malicious purposes, including the creation or dissemination of deceptive or harmful media, is strictly prohibited. Access requests must include a clear research purpose, and the authors reserve the right to deny access at their discretion. Extraction Instructions:After downloading from Zenodo, extract each .zip archive to access the videos in its respective category. Example: unzip deepfake_face_swapped_1.zip Each .zip file is self-contained and does not require merging with other archives.标题：DeepFakeX：面向深度伪造研究的综合数据集 DeepFakeX是一套精心甄选的数据集，涵盖5个类别共计800段视频： - 人脸替换型深度伪造视频（500段） - 音频替换型深度伪造视频（100段） - 语音克隆型深度伪造视频（100段） - 人脸与音频联合替换型深度伪造视频（100段） 本数据集采用多款当前前沿的深度伪造制作工具生成，以保障样本的高多样性与真实度；同时剔除了政治人物、未成年人及其他敏感个人素材，以降低隐私泄露与恶意滥用风险。 每段视频均带有对应篡改类型的专属前缀标签（ff_、fa_、vc_、ffa_），用于标识其深度伪造类别。数据集按类别打包为压缩归档文件（例如：deepfake_face_swapped_1.zip、deepfake_voice_cloned.zip）。特别地，数据集附带CSV格式元数据文件，可为每个视频样本提供结构化信息，涵盖视频标识符、文件名、深度伪造类别、生成方法、语言、场景上下文、分辨率及时长。 许可证：知识共享署名-非商业性使用4.0国际许可协议（CC BY-NC 4.0，Creative Commons Attribution–NonCommercial 4.0 International） 免责声明：本数据集仅可用于深度伪造检测与多媒体取证领域的非商业学术研究。任何用于恶意目的的使用行为（包括制作或传播欺骗性、有害媒体）均被严格禁止。申请访问数据集需明确说明研究用途，作者保留自行决定拒绝访问申请的权利。 解压说明：从Zenodo平台下载数据集后，解压各分类对应的.zip归档文件即可获取对应类别的视频。示例命令： unzip deepfake_face_swapped_1.zip 每个.zip文件均为独立完整的归档，无需与其他归档文件合并。