Publication

Abstract : Recent technologies have evolved rapidly, allowing the creation of fake videos/images that are realistic. These artificial media pieces have provoked grave worries about the possibilities of their misuse in different areas like politics, journalism, and communication. Consequently, the demand for deepfake detection methods is increasing. In response to this growing threat, there is an increasing demand for robust deepfake detection methods. This paper proposes an ensemble-based approach for deepfake detection, leveraging deep learning techniques. Specifically, we employ a combination of CNN models, including VGG, Xception, RegularizedConvNet, and RegularizedConvDenseNet, designed for image classification. Our approach integrates predictions from multiple models to enhance detection accuracy and reliability. Specifically, the base is formed by the ensemble of RegularizedConvNet and RegularizedConvDenseNet, which cooperate to prevent the spread of deepfakes and protect digital integrity. Experimental results show that our Ensemble method achieve better detection values with accuracy of 94.5% and AUC of 0.98.