Abstract

The growth of deep fake technology, driven by generative adversarial networks (GANs), has increased the production of manipulated and fake videos, causing several serious threats to digital authenticity, privacy, and societal trust. This research introduces a Convolutional Neural Network (CNN) based framework designed to detect fake videos by analyzing different extracted frames from a custom organized dataset. Leveraging OpenCV for preprocessing and a tailored CNN architecture, our approach achieves an impressive accuracy of 90% while providing a lightweight yet effective alternative to resource-heavy models like ResNet and Vision Transformers (ViTs). This study compares our method with state-of-the-art techniques in terms of efficiency and scalability. By addressing the increasing challenge of fake video creation through artificial intelligence, this work contributes to video forensics and lays the groundwork for real-time detection systems that can effectively overcome the threat of fake videos and misinformation.