With the widespread availability of powerful editing tools in today’s digital environment, image forgeries have become a significant concern across various fields. Conventional detection techniques frequently fall short when faced with complex forgeries that employ splicing or copy-move techniques. We use an ensemble transfer learning approach with pre-trained VGG-16, VGG-19, and XceptionNet architectures on large datasets to address this. These models excelled at extracting robust and discriminative features for image forgery detection. The ensemble approach harnessed the strengths of each model, capturing complementary features and mitigating the limitations of relying on a single architecture. The extracted feature representations from the ensemble are then fed into a Support Vector Machine (SVM) classifier, renowned for its ability to separate data points in high-dimensional spaces. This potent combination of SVM and transfer learning ensembles made effective forgery detection possible, which took advantage of SVM’s strong classification performance on the difficult forgery task and deep neural networks’ rich representational capabilities.