Designing and Implementation of Image Forgery Detection Algorithm

Authors:

Nikita Tembhurne¹, Prajkta Girhe², Sharwari Choudhari³, Rani Sawarkar^4, Anil Warbhe⁵

¹Dr. Anil Warbhe, Information Technology & Nagpur Institute of Technology

²Sharwari Chaudhari Information Technology & Nagpur Institute of Technology

Abstract - The rapid evolution of Generative Adversarial Networks (GANs) has enabled the production of highly realistic synthetic images, raising serious concerns about digital media authenticity and contributing to misinformation and deepfake proliferation. This project aims to design and implement a digital image forensics algorithm capable of accurately detecting GAN-generated images. It begins with a comprehensive review of existing detection approaches, including CNN-based classifiers, frequency domain analysis, and artifact-based methods, highlighting their limitations such as poor generalization to unseen GAN models and susceptibility to anti-forensic techniques. To address these challenges, the project proposes a hybrid detection framework that combines attention-based convolutional neural networks with frequency-domain feature extraction to improve robustness and accuracy. Transfer learning and ensemble techniques are incorporated to reduce overfitting and enhance the detection of subtle inconsistencies like spectral irregularities and semantic distortions

Key Words:  Digital Image Forensics, Image Forgery Detection, GAN Image Detection, Generative Adversarial Networks, Image Authenticity, Deepfake Detection, Synthetic Media, Artifact Detection