Image Forgery Detection Using Deep Learning

Abstract

Image forgery is a big problem in digital media, making it important to have strong detection methods to fight misinformation and keep trust in visual content. In this project, we introduce an advanced image forgery detection system using VGG16, a powerful convolutional neural network, and Error Level Analysis (ELA) algorithms. Our goal is to create an efficient and accurate system that can identify real images from fake ones, especially focusing on detecting splicing and copy-move forgeries. By examining pixel intensities and patterns, our system can accurately find tampered areas, improving the integrity and trustworthiness of digital images. We use a diverse dataset of real and fake images from different sources to train and test the VGG16-ELA model. We aim to find the percentage of forgery, highlighting the forged areas and generating the mask of forged area. Through this effort, we aim to increase trust in visual content in fields like forensics, journalism, and social media, helping to ensure the reliability of digital information.