Deep Fake Video Detection through Deep Learning

Abstract

In the last century, rapid development in AI, machine learning, and deep learning has inspired new methodologies and a lot of tools specifically designed for the treatment of multimedia content. This type of video is created based on the replacement of face features of one person using sophisticated Deep Learning methods that are used to replace others. Although the technology is powerful, the risk lies in its possible exploitation for misinformation, manipulation, and deceptive persuasion. Even with ongoing research efforts, there is still a lack of effective solutions for reliably detecting deep fake content. Nonetheless, great strides are being made in addressing this challenge through dedicated research endeavours. The issue of deep fakes makes distinguishing between real and fake videos a challenge. The human eye cannot identify the difference directly because deep fakes are highly realistic. Identifying these videos is challenging because of advancements in increasingly authentic deep fake creation technologies that periodically emerge. Various approaches have been proposed in the literature to tackle the challenges posed by deep fakes. In this paper, we present an updated overview of deep fake detection through comprehensive research efforts after a literature survey. The proposed model depends on the transfer learning basis of convolutional neural network, wherein the model will train to detect the noise embedded inside deep fakes. In this paper, we will be discussing current deep fake detection strategies incorporating conventional as well as the advanced techniques of learning. Deep learning emerges as one of the viable approaches The proposed model classified that accuracy of fake videos against real ones was to achieve 98.0 %. Besides, we examine and discuss the efficiency of varied techniques used in Deep fake content detection over diverse datasets. We conclude that, at all times, Deep learning-based methods prove brilliant over other approaches while applying to Deep fake detection techniques.