Fake image detection using deep learning

Abstract

Detecting digital image forgery has become a major challenge in the field of cybersecurity and visual content verification. This thesis presents a deep learning-based system for detecting and localizing Copy-Move forgeries in images. The proposed model is built on a hybrid architecture that combines Convolutional Neural Networks (CNNs) for local feature extraction and Transformers for capturing global context. To enhance performance, we integrated Focal Loss and the Mixup data augmentation technique during training. The system was evaluated on the CoMoFoD dataset, demonstrating a high ability to detect tampered images and accurately highlight the forged regions. Additionally, a user-friendly local desktop application was developed, allowing users to upload an image, receive a prediction, and clearly visualize the suspected manipulated zones.