Published Paper Details:

Image manipulation and forgery present significant challenges across various domains, including forensics, security, and media authentication. This is a novel approach that integrates Error Level Analysis (ELA) with Convolutional Neural Networks (CNNs) to differentiate between authentic and tampered images. Using the Cassia v2 dataset containing both real and fake instances, diverse data augmentation techniques, such as flattening, resizing, and converting images to ELA format, are applied to enhance model robustness. The dataset is partitioned into 80% training and 20% validation sets to facilitate comprehensive model training and evaluation. Utilizing Keras, a Sequential model is developed, incorporating Conv2D, MaxPooling2D, Dropout, Flatten, and Dense layers for effective feature extraction and classification. Training utilizes the Adam optimizer for parameter optimization. Evaluation metrics, including loss, accuracy, and a confusion matrix, are employed to assess model performance. Results demonstrate promising accuracy, with the model achieving 98.8% training and 92.8% validation accuracy, showcasing the efficacy of the proposed methodology in accurately distinguishing between real and fake images. This approach holds potential for applications in image forensics, security, and authentication domains.