华中科技大学学报（自然科学版）2025，Vol.53Issue(5)：85-103,19.DOI:10.13245/j.hust.250021
深度伪造人脸生成与检测技术综述
Survey of deepfake face generation and detection technologies
摘要
Abstract
In the face of the serious threat posed by the application of deepfake technology in the field of face forgery to information security,the latest progress and main characteristics of deepfake face generation technology were first comprehensively reviewed and summarized.This technology has strong deceptiveness,low forgery cost,and high detection difficulty,making it difficult for the public and existing detection methods to effectively distinguish and detect.Secondly,according to the differences in forgery types,deepfake face generation technology was divided into four categories,which were complete face generation,attribute editing,identity replacement,and facial reenactment.For each type of technology,a detailed elaboration and analysis were carried out to clarify its technical principles and application scenarios.Thirdly,the datasets of real faces and deepfake faces involved in deepfake face detection technology were systematically summarized.Meanwhile,starting from feature selection,the existing deepfake face detection methods were classified,and a detailed analysis and comparison were made,including detection methods based on biometric features,identity information,image spatial features,image frequency-domain features,temporal features,and hybrid features.Finally,the challenges and future research directions in the fields of deepfake face generation and detection technology were deeply explored,respectively.关键词
深度伪造技术/人脸伪造/深度伪造检测/媒体取证/信息安全Key words
deepfake technology/facial forgery/deepfake detection/media forensics/information security分类
信息技术与安全科学引用本文复制引用
杨宏宇,李星航,胡泽..深度伪造人脸生成与检测技术综述[J].华中科技大学学报（自然科学版）,2025,53(5):85-103,19.基金项目
国家自然科学基金民航联合研究基金重点项目(U2433205) （U2433205）
江苏省基础研究计划自然科学基金青年基金资助项目(BK20230558). （BK20230558）
