As digital tools have advanced, image manipulation has become more common, resulting in various forgeries that mask original content. Detecting manipulated regions is more difficult by methods like copy-pasting or image merging, especially when geometric changes are involved. Using three crucial steps—pre-processing, image augmentation, and classification—this study presents a novel deep learning-based method for identifying image forgeries in digital images. Image normalization, rescaling, and error level analysis (ELA) are pre-processing methods that reduce overfitting and improve model accuracy. Image augmentation is used to increase the dataset size and ensure the model is exposed to a greater range of training samples to increase robustness. The suggested approach, implemented on the Python platform, uses convolutional transfer learning for classification. The Columbia, CASIA V1.0, and CASIA V2.0 datasets are used to assess it. The experimental results demonstrate outstanding performance with an accuracy of 99.97%, precision of 95.63%, recall of 98.95%, and an F1 score of 97.26%. These metrics highlight the method’s potential for real-world applications in image integrity verification by demonstrating its superior efficacy in detecting forgeries compared to traditional techniques.