Deepfake technology is causing unprecedented threats to the authenticity of digital media, and demand is high for reliable digital media detection systems. This systematic review focuses on an analysis of deepfake detection methods using deep learning approaches, machine learning methods, and the classical methods of image processing from 2018 to 2025 with a specific focus on the trade-off between accuracy, computing efficiency, and cross-dataset generalization. Through lavish analysis of a robust peer-reviewed studies using three benchmark data sets (FaceForensics++, DFDC, Celeb-DF) we expose important truths to bring some of the field’s prevailing assumptions into question. Our analysis produces three important results that radically change the understanding of detection abilities and limitations. Transformer-based architectures have significantly better cross-dataset generalization (11.33% performance decline) than CNN-based (more than 15% decline), at the expense of computation (3–5× more). To the contrary, there is no strong reason to assume the superiority of deep learning, and the performance of traditional machine learning methods (in our case, Random Forest) is quite comparable (accuracy of 99.64% on the DFDC) with dramatically lower computing needs, which opens up the prospects for their application in resource-constrained deployment scenarios. Most critically, we demonstrate deterioration of performance (10–15% on average) systematically across all methodological classes and we provide empirical support for the fact that current detection systems are, to a high degree, learning dataset specific compression artifacts, rather than deepfake characteristics that are generalizable. These results highlight the importance of moving from an accuracy-focused evaluation approach toward more comprehensive evaluation approaches that balance either generalization capability, computational feasibility, or practical deployment constraints, and therefore further direct future research efforts towards designing systems for detection that could be deployed in practical applications.

The development of deepfake technologies is a threat to digital media authentication and cybersecurity infrastructure. The current paper proposes a method for detecting manipulated images of faces based on the Vision Transformer architecture. We fine-tune a pre-trained ViT-Base-Patch16-224 model based on this well-curated dataset of 12,137 face images, which includes an almost equal number of real and synthetic face images using a variety of different generation methods. The data set contains real-life photographs of CelebA and FFHQ, along with artificial samples of the publicly available Kaggle repositories (FaceForensics++, Celeb-DF, and DFDC) and 600 self-collected photos (300 real-life photographs of personal cell phones and 300 artificial ones created with the help of modern tools) to make it closer to real-life use. Methodology involves assessment of the quality of data, systematic preprocessing by ImageNet normalization, data augmentation, and stratification of a 70-20-10 partitioning of the data. AdamW optimization using a learning rate of 2 × 10–5 was used with 8 epochs. The test accuracy of the system was 99.01%, the precision was 98.85%, and the recall was 99.18%, with 12 misclassifications. The results demonstrate that Vision Transformers can effectively model global image dependencies for detecting Deepfakes. The complete methodology is documented in this paper to ensure reproducibility.