Work

Deepfake face Detection using Deep Learning Technique: A Symmetric Review

Year: 2026
Type: article
Abstract: The rapid advancement of artificial intelligence and deep learning has enabled the creation of highly realistic synthetic facial content, commonly known as deepfakes. Although deepfake technology has beneficial applications in entertainment, education, and digital media, its misuse poses serious challenges related to misinformation, identity theft, privacy, and digital security. Deepfake face detection has therefore emerged as an important research area in computer vision and multimedia forensics. This symmetric review provides a comprehensive analysis of deep learning-based approaches developed for detecting manipulated facial images and videos. The review examines conventional Convolutional Neural Networks (CNNs), transfer-learning models, recurrent architectures, and advanced networks such as ResNet, DenseNet, EfficientNet, and Vision Transformer-based approaches. It systematically compares these techniques in terms of detection accuracy, robustness, computational complexity, dataset requirements, and generalization capability. Particular attention is given to facial preprocessing, frame extraction, feature representation, spatial and temporal artifact analysis, data augmentation, and evaluation metrics. The review also discusses widely used deepfake datasets and highlights challenges associated with compression, unseen manipulation techniques, cross-dataset performance, and adversarial attacks. Finally, emerging research directions, including multimodal learning, explainable artificial intelligence, lightweight detection models, and hybrid spatial-temporal architectures, are discussed. The study provides a structured perspective on the current state of deepfake face detection and identifies promising opportunities for developing more reliable, generalizable, and real-time detection systems.
Language: English
Open Access status: hybrid
OpenAlex ID: W7217830426
FWCI:
Cites: 16
Cited by:
Related to: 10