Image Authentication: Differentiating Camera-Captured and AI-Generated Images with Provenance Verification using Vision Transformer (ViT)

The differential impact of image reconstruction: Generative model artifacts as forensic signatures

A Hierarchical ELA-CNN Framework for Image Forgery Detection and Manipulation Localization

Increasing availability of advanced AI-based image creation and editing tools has significantly increased the prevalence of digital image forgery, creating challenges for media authenticity, digital forensics, and information security. This paper presents an intelligent image forgery detection framework that combines multi-quality Error Level Analysis (ELA) with an optimized Convolutional Neural Network (CNN) and heat map-based manipulation localization. The proposed approach utilizes ELA preprocessing at multiple JPEG quality levels to amplify compression inconsistencies associated with image tampering, this is followed by a lightweight CNN architecture trained using advanced optimization techniques, including warmup and cosine annealing learning-rate scheduling, stochastic gradient descent with momentum, and class imbalance handling. To improve interpretability, a heat map generation module is introduced to localize suspicious image regions and provide visual evidence supporting classification decisions which provides interpretability and explainability. The framework was evaluated using the CASIA 2.0 image forgery dataset containing a total of 12,614 images, 7,491 authentic and 5,123 tampered images. Experimental results demonstrate an overall classification accuracy of 94%, outperforming a baseline ELA-CNN implementation while maintaining computational efficiency. The proposed localization mechanism enhances explainability by highlighting potential manipulation regions. A web-based interface and RESTful API were developed to support practical deployment and integration into digital forensic workflows. The results indicate that the proposed framework provides an effective and interpretable smart solution for automated image forgery detection and localization.

Explainable Deep Learning Framework for AI-Generated Image Detection Using NASNet and Grad-CAM

The rapid growth of generative artificial intelligence has made it easier to create highly realistic synthetic images, increasing the risk of misinformation, identity misuse, and digital fraud. Distinguishing AI-generated images from authentic ones has become a significant challenge due to their visual similarity. This project presents an explainable deep learning framework for identifying real and AI-generated images using the NASNet architecture. The model is trained on a balanced dataset containing genuine and synthetic face images after applying preprocessing techniques such as resizing, normalization, and data shuffling. To improve transparency, the system integrates Explainable AI (XAI) methods, including Grad-CAM and LIME, which highlight the image regions that influence the model’s predictions. A web-based interface enables users to upload images in different formats and receive instant classification results. Experimental evaluation demonstrates that the proposed approach achieves high detection accuracy while providing interpretable visual explanations, making it suitable for digital image verification, media authentication, and cybersecurity applications.

Deep Learning-Based Fake Image Detection Using Transfer Learning: A Systematic Review

The rapid advancement of artificial intelligence (AI) has significantly changed the way digital visual content is created, enabling the generation of highly realistic synthetic images and videos. While these technologies support many beneficial applications, they have also facilitated the creation of manipulated visual content, commonly known as deepfakes, which pose serious challenges to information authenticity, public trust, cybersecurity, and digital forensic investigations. As image manipulation techniques continue to evolve through advanced models such as Generative Adversarial Networks (GANs) and diffusion-based frameworks, conventional detection methods relying on handcrafted features have become increasingly inadequate. In response, deep learning approaches integrated with transfer learning have emerged as effective solutions due to their ability to leverage pre-trained models for extracting robust and discriminative features, even when limited training data are available. This review presents a comprehensive analysis of recent deep learning and transfer learning techniques for fake image detection. It examines widely adopted convolutional neural network (CNN) architectures, benchmark datasets, evaluation metrics, and current research developments. Furthermore, the paper provides a comparative assessment of existing methods by highlighting their strengths, limitations, and performance characteristics. Finally, it identifies major research challenges and outlines future directions for developing robust, scalable, and generalizable fake image detection systems capable of addressing the growing threats posed by AI-generated visual content in cyberspace

Integration of Invisible Watermark Using Hybrid DWT-SVD in an AI Image Generator for Content Validation

Training-Free Reconstruction-Based AI-Generated Image Detectors Are Inherently Vulnerable to Adversarial Examples

The impressive visual quality and ubiquity of AI-generated images call for reliable and robust detection methods. Reconstruction-based detectors have emerged as a promising direction for transparent and training-free identification of synthetic images. However, due to their fundamentally different mode of operation (compared to standard, classifier-based methods), little is known about their adversarial robustness. In this work, we propose two novel attack methods targeted at detectors that leverage autoencoder reconstruction error. We find that by constructing imperceptible adversarial examples, the distance between original and reconstruction can be artificially increased, causing fake images to be wrongly classified as real. Our evaluation including images from three state-of-the-art generators and three detectors demonstrates that detection performance is significantly decreased, even if attacked images additionally undergo real-world degradations. Critically, our adversarial examples naturally transfer across detectors, as they all share the same principle, pointing towards an inherent vulnerability of reconstruction-based detectors.