A CNN Algorithm Based DeepFake Detection System

DeepFake is a technology that has garnered significant attention in the modern world and is often used for unlawful and immoral practices. The consequences and aftereffects of generation and publication of deepfakes in the public domain can tamper the social image of organizations or individuals. The objective of this study is to find a solution to this upcoming societal horror which can ruin lives and stigmatize the affected individuals. This project mainly concentrated on the development of a DeepFake Detection model that leverages a custom Convolutional Neural Network (CNN) to identify facemanipulated images or video content by extracting frames from images or videos. The model achieved high accuracy in demonstration of effectiveness of custom CNN architecture in identification of subtle facial inconsistencies and performed robustly across the dataset which highlighted the capability for generalization and detection of face-manipulated content in images and videos, alike. In essence, this CNN model stands out as a comprehensive solution not only addressing and amending the challenges faced by current systems but also enables people to be alert. This research serves as a foundational step towards the development of robust and scalable solutions to effectively counter DeepFake technology.