Ensemble of Machine Learning Classifiers for Detecting Deepfake Videos using Deep Feature

Rapid progress in artificial intelligence, machine learning, and deep learning over the last few decades has resulted in new methodologies and tools for altering multimedia. Deepfakes represent a technology for seamlessly swapping faces in videos, yielding remarkably lifelike outcomes. While this technology offers numerous applications, its malicious use, such as disseminating false information or participating in cyber harassment, can exert a significant impact on society. This makes the identification of deepfakes a critical issue. In this paper, we propose a hybrid strategy for deepfake identification in videos that combines deep learning and machine learning. Faces are identified in the videos using YOLO-V3 face detectors and using the efficientNet deep learning model, features are extracted from the faces. Deepfakes are identified using an ensemble of machine learning classifiers such as support vector machine (SVM), decision trees(DT), k-nearest neighbor (KNN), and naive bayes(NB) based on the max voting approach, which provides better results for datasets of varying sizes and resolutions. Experiments are carried out by integrating the Celeb-DF(v2) and FaceForensics++ (FF++) datasets and the suggested technique achieves 99.64% accuracy and proves that the suggested method is more effective than state-of-the-art methods.