Abstract—Deepfake technology has emerged as a major challenge in combating misinformation and manipulated digital content, posing serious risks to individual privacy, institutional credibility, and public trust. The increasing accessibility of generative models has significantly amplified the production and dissemination of highly realistic synthetic media, thereby intensifying the need for robust detection mechanisms. This study investigates facial data augmentation techniques employed to enhance machine learning and deep learning models for Deepfake detection. It presents a structured and comprehensive survey of augmentation strategies, highlighting their methodological foundations, practical implementations, and reported effectiveness. The review systematically examines geometric and photometric transformations, noise injection methods, occlusion strategies, cutout and mixing-based augmentations, and commonly used Python-based augmentation libraries, assessing their contribution to improving model generalization and mitigating overfitting. In addition, leading Deepfake detection approaches are summarized alongside the datasets and performance metrics frequently reported in the literature, including accuracy and Area Under the Curve (AUC), to provide a clearer understanding of existing evaluation practices. The study further discusses key challenges, dataset dependencies, and unresolved research issues that influence augmentation effectiveness across different experimental settings. By synthesizing current findings and identifying methodological limitations, this review aims to guide researchers in selecting and designing augmentation strategies that support the development of more robust, reliable, and generalizable Deepfake detection frameworks.
 
Keywords—deepfake detection, machine learning, data augmentation, image preprocessing
 
Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).