Deepfake Image Detection Using Convolutional Neural Network

Abstract

The recent advancements in the field of Artificial Intelligence and Machine Learning allow the unethical practice of generation of forged videos by using face swaps and imitating the facial expressions, voice, features and other attributes to serve malicious purposes. The difficulty in distinguishing between the real and the forged videos poses a serious threat to the society, democracy and privacy of a person. The remarkable quality of the forged videos along with the widespread reach through social media platforms manipulates public opinion by influencing their sentiments and perceptions. It is therefore essential to devise ways through which deepfake videos can be detected because otherwise, its repercussions are devastating beyond one’s imagination. This paper firstly provides a comprehensive overview of the various deepfake generation and detection techniques andthenpresents a deepfake detection model using Inception-v3 Architecture.