Unmasking Deepfakes: CNNs and Vision Transformers for Cutting-Edge Detection
Year: 2025
Type: conference-paper
Abstract: Deepfake technology, driven by Generative Adversarial Networks (GANs), poses challenges in digital security by enabling highly realistic synthetic media. This study proposes a detection framework combining Convolutional Neural Networks (CNNs) for local feature extraction and Vision Transformers (ViTs) for global analysis. Evaluated on two datasets, CNNs achieved 97.1 % accuracy on a140K-image dataset, outperforming ViTs at 90.06%, though ViTs showed better generalization. Despite these advances, deepfake detection faces challenges like adversarial attacks and dataset biases. Future work will enhance real-time processing, robustness, and multi-modal approaches integratingaudio and behavioral cues.
Authors: SAADI, Abdelhalim, SLIMANI, Yacine, Louze, Ridha, Hammadou, Roufaida
Institutions: University Ferhat Abbas of Setif, Université Constantine 2
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7104370716
FWCI:
Cites:
Cited by:
Related to: