Multi-scale Image Tampering Detection Using Inception-UNet Network

Image tampering detection techniques are being needed in many fields. For example, in the field of forensic evidence, there is a lack of reliable techniques for verifying the authenticity of digital image evidence. In the academic field, qualified detection techniques are also required to determine the authenticity of images in research papers. Therefore, image tampering detection techniques can significantly reduce human resource consumption. In this work, a network using a combination of the Inception module and U-Net is proposed, which extracts the multi-scale features of the image for tampering detection. This method extracts the multi-scale features of the image for tampering detection using Inception module. The feature information is processed through the U-network and residual structure, and the detection result of the image is output after up-sampling step by step. The method also extracts image noise using constrained convolution operation. Invalid features are suppressed through the attention mechanism, which ultimately leads to good prediction of image tampering. It is experimentally verified that our method has good prediction performance for tampered images.