Deepfake generation and detection, a survey

• Published in: Multimedia tools and applications Vol. 81; no. 5; pp. 6259 - 6276
• Main Author: Zhang, Tao
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.02.2022; Springer Nature B.V
• Subjects: Acoustics; Computer Communication Networks; Computer Science; Data Structures and Information Theory; Datasets; Deception; Deepfake; Image manipulation; Machine learning; Multimedia Information Systems; National security; Special Purpose and Application-Based Systems; Deepfake; Generation; Survey; Detection; Media forensics
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-021-11733-y

Deepfake refers to realistic, but fake images, sounds, and videos generated by articial intelligence methods. Recent advances in deepfake generation make deepfake more realistic and easier to make. Deepfake has been a signicant threat to national security, democracy, society, and our privacy, which calls for deepfake detection methods to combat potential threats. In the paper, we make a survey on state-ofthe-art deepfake generation methods, detection methods, and existing datasets. Current deepfake generation methods can be classified into face swapping and facial reenactment. Deepfake detection methods are mainly based features and machine learning methods. There are still some challenges for deepfake detection, such as progress on deepfake generation, lack of high quality datasets and benchmark. Future trends on deepfake detection can be efficient, robust and systematical detection methods and high quality datasets.