| 人脸视频深度伪造与防御技术综述 |
| |
| 引用本文： | 周文柏,张卫明,俞能海,赵汉卿,刘泓谷,韦天一. 人脸视频深度伪造与防御技术综述[J]. 信号处理, 2021, 37(12): 2338-2355. DOI: 10.16798/j.issn.1003-0530.2021.12.007 |
| |
| 作者姓名： | 周文柏 张卫明 俞能海 赵汉卿 刘泓谷 韦天一 |
| |
| 作者单位： | 中国科学技术大学,网络空间安全学院 |
| |
| 基金项目： | 国家自然科学基金（U20B2047, 62002334）;安徽省自然科学基金（2008085QF296）;中国科学技术大探索类基金（YD3480002001）以及中国科学技术大学青年基金（WK2100000011）资助
|
| |
| 摘 要： | 近年来,得益于深度生成模型的发展,人脸的操控技术取得了巨大突破,以Deepfake为代表的人脸视频深度伪造技术在互联网快速流行,受到了学术界和工业界的广泛重视.这种深度伪造技术通过交换原始人脸和目标人脸的身份信息或编辑目标人脸的属性信息来合成虚假的人脸视频.人脸深度伪造技术激发了很多相关的娱乐应用,如使用面部替换技术将...
|
| 关 键 词： | 人脸视频深度伪造 人脸伪造视频防御 生成技术 视频取证 检测技术 主动防御 |
| 收稿时间： | 2021-07-01 |
An Overview of Deepfake Forgery and Defense Techniques |
| |
| Affiliation: | University of Science and Technology of China, School of Cyberspace Science and Technology |
| |
| Abstract: | Benefiting from the development of the deep generative model, face manipulation technology has made great breakthroughs. The deep face forgery technology, represented by Deepfake, has rapidly become popular on the Internet and received extensive attention from academia and industry. This deep forgery technique synthesizes fake face videos by exchanging the identity information between original and target faces or by editing the attribute information of target faces. The deep face forgery technique has inspired many related entertainment applications, such as using facial substitution to replace the user's face in a movie clip, or using expression reenactment to drive a static portrait of a famous person. However, the current deep face forgery technology is still in the fast-growing stage, and its authenticity and naturality require further improvement. On the other hand, this kind of deep face forgery can easily be used by malicious criminals to produce pornographic movies, fake news, or even political rumors about important dignitaries, which is a great potential threat to national security and social stability, so the defense technology of face forgery videos is crucial. In order to reduce the negative impact of deep face forgery videos, many researchers have investigated the detection and identification techniques of face forgery videos, proposing a series of defense methods from different perspectives. Unfortunately, due to the single form of dataset distribution, inconsistent evaluation metrics and lack of initiative, etc., there is still a long way to go before the detection technology can be applied practically. In fact, the research on deep face forgery and detection techniques remains in the developmental stage, their connotations and extensions are being rapidly updated and iterated. In this review, we will make a scientific and systematic summary of the main research work so far, and briefly analyze the limitations of existing technologies. Finally, we will discuss the potential challenges and development directions of deep face forgery and detection technology, in order to provide insights for future research work in the field. |
| |
| Keywords: | |
|
| 点击此处可从《信号处理》浏览原始摘要信息 |
|
点击此处可从《信号处理》下载免费的PDF全文 |
|
