Improved Deep Learning Based-Model for Fake Image Detection

The rapid proliferation of digitally fake images presents a significant obstacle to the veracity of the visual information. As the technology for altering images has advanced, it has become increasingly difficult to distinguish between real and fake content. The aim of this study is to tackle the pressing issue of effectively detecting fake images by utilizing convolutional neural networks (CNNs) and the error-level analysis (ELA) method. Our main goal was to establish a reliable approach that can precisely detect fake images, thereby proving the capabilities of digital forensics. We used CASIA Version 2 dataset to assess the effectiveness of the proposed method. Because of its diversity and complexity, this dataset offers a challenging testing ground for the evaluation of fake image detection methods. The obtained results demonstrate the ability of the proposed approach to accurately identify fake images. Therefore, an accuracy of 93.04% was achieved using the dataset.