虚假数字人脸内容生成与检测技术

In recent years, artificial intelligence technologies, represented by deep learning algorithms, have been widely applied in many fields including smart video surveillance, privacy protection, autonomous driving and so on. Particularly in the field of face recognition, deep learning based methods have shown the ability of surpassing human perception and brought great conveniences to our daily lives. However, digitally forged face generated by adversarial and deepfakc techniques poses huge risks and chalenges for individual privacy,social security and even nation-alsecurity. This paper reviews previous workonthe creation and detection of digitaly forged face content,revealing potential risks to individual privacy, social security and national security . Specificaly,we firstly introduce the atack targets and atack types of digitaly forged face content. Secondly, we summarize and analyze digitaly forged face content creation, atack, detection and defense technologies in terms of two atack targets: artificial inteligence system and human perception system, and in terms of two atack types: adversarial face example and deepface manipulation.Finaly,the directions of future research on digitaly forged face content creation and detection are discussed.