Robust Image Forgery Detection and Localization Framework using Vision Transformers (ViTs)

Abstract

Image forgery detection has become increasingly critical with the proliferation of image editing tools capable of generating realistic forgeries. Traditional deep learning approaches, such as convolutional neural networks (CNNs), often struggle with capturing global dependencies and subtle inconsistencies across larger image contexts. To address these challenges, this paper proposes a novel Vision Transformer(ViT)- based framework for robust image forgery detection and localization. Leveraging the self-attention mechanism of transformers, our approach effectively models long-range dependencies and detects even subtle tampered regions with high precision. The proposed framework processes images as patch embeddings, extracting both local and global features, and outputs a detailed forgery map for accurate localization. We evaluate our method on multiple benchmark datasets containing diverse forgery types, including splicing, cloning, and inpainting. Experimental results demonstrate that the ViTbased model outperforms state-of-the-art CNN and GAN-based methods, achieving superior accuracy, precision, and recall. Additionally, qualitative analyses highlight its capability to localize forgeries in complex scenarios. The results underscore the potential of Vision Transformers as a powerful tool for advancing the fieldof image forgery detection.