Title: Deepfake detection based on single-domain data augmentation

Authors: Qian Feng; Zhifeng Xu

Addresses: College of Cyber Security, Jinan University, Guangzhou, 510632, Guangdong, China ' College of Cyber Security, Jinan University, Guangzhou, 510632, Guangdong, China

Abstract: Deepfake has posed a serious threat to personal privacy and social stability. The related research on deepfake detection (DFD) has gained sufficient high accuracy on various datasets, while the generalisation performance is still insufficient. Most of the existing methods are aimed at analysing and detecting specific traces and distortions generated by a specific forgery algorithm. However, these detection algorithms typically experience a significant decline in accuracy when detecting forgery videos generated by other algorithms. This paper proposed a Deepfake detection scheme based on Single-Domain Data Augmentation, and considered the most difficult situation in the DFD generalisation problem: How to generalise to a variety of unknown forgery data when only the real data is known. We proposed the universal forgery generation (UFG) and adversarial style transfer algorithm (AST) to augment forgery data and improve generalisation ability. The experimental results show that our scheme can simulate diverse unknown forgery data, and the generalisation performance is superior to many existing schemes.

Keywords: adversarial training; data augmentation; deepfake detection; domain generalisation; style transfer.

DOI: 10.1504/IJAACS.2025.148530

International Journal of Autonomous and Adaptive Communications Systems, 2025 Vol.18 No.4, pp.293 - 309

Received: 14 Apr 2024
Accepted: 04 May 2024
Published online: 11 Sep 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article