De-faketection: Deepfake detection via convolutional neural network (DFT)

Abstract

Deepfake is a technology that forges videos and images by substituting faces. Over the years, it has gradually improved and managed to produce realistic deepfakes that are difficult to be determined even with the trained eyes. With this, the problem of detecting deepfakes has received the public's attention and therefore, this project is initialized to develop a deepfake detection tool with the integration of machine learning algorithms - Convolutional Neural Network (CNN). It is a neural network particularly used to categorize features of real and fake graphical input. Apart from that, various detection algorithms have been studied and concluded that most proposed algorithms are difficult to be performed by general users. Therefore, this project is initialized to allow users without any prior knowledge to detect deepfakes through the development of a simple yet functional detection tool.