Effectiveness of deep learning architecture for pixel-based image forgery detection

Digital image forgery or forgery is easy to do nowadays. Verification of the authenticity of images is important to protect the integrity of the images from being misused. The use of a deep learning approach is state-of-the-art in solving cases of pattern recognition, the one is image data classification. In this study, image forgery detection was carried out using a deep learning-based method, the Convolutional Neural Network (CNN). The analysis of the different architecture of CNN has been done to show the effectiveness of each architecture. Two architectures were tested to know which one is more effective, architecture 1 has three convolution and pooling layers with 256 × 256 × 3 image input. While the other architecture has two convolution layers and pooling with 128 × 128 × 3 image input. The results show that the accuracy rate of the image forgery detection model in each architecture is around 80%. However, the validation accuracy is not more than 70%