Detecting Deepfake Human Face Images Using Transfer Learning: A Comparative Study

Deepfake technology has made it easy to create realistic, manipulated videos and images, raising concerns about the potential for them to be used to spread misinformation or harm individuals. Deepfake technology has advanced significantly in recent years, making it increasingly difficult to distinguish between real and fake images. This has raised concerns about the potential for deepfake images to be used maliciously, for example, in political propaganda or cyber fraud. As a result, there is a growing need for effective methods of detecting deepfake images. So, it became important to propose an effective classification on deepfake. In this paper the authors have proposed an effective classification on deepfake images using Transfer Learning. This approach will help in classification of Real and fake through Human face images. The authors have utilized Real vs Fake Face classification dataset to apply the proposed approach. The Loss, Accuracy, Precision, Recall, AUC, and F1 Score parameters have been used to analyze the performance of proposed approaches. The outcome of this work shows the EfficientNetV2L Model with the accuracy of 99% as compared among EfficientNetV2B0-B3 and S, M, L models.