Deepfake Creation and Detection using Ensemble Deep Learning Models

Abstract

The use of Artificial Intelligence to create falsified videos using Deep Neural Networks is posing a serious problem in distinguishing the real from the counterfeit. These counterfeit videos are known as "Deepfakes". Due to their realistic appearance and their subsequent ability to influence perceptions and mass sentiment, deepfakes must be monitored. Malicious deepfakes must be detected, and their circulation is immediately controlled. Many deepfake detection technologies have been developed that use particular features to classify fabricated media. This paper proposes the framework of deepfake detection using deep neural network models. The hybrid combination of deep learning models predicts deepfakes with better accuracy. The proposed model is tested and evaluated on the DFDC and CelebDF dataset that classifies more deepfake videos. © 2022 ACM.