MSRD-CNN: Multi-Scale Residual Deep CNN for General-Purpose Image Manipulation Detection

The authenticity of digital images is a major concern in multimedia forensics due to the availability advanced photo editing tools/devices. In literature, several image forensic methods are available detect specific processing or operations. However, it remains challenging task design universal method that can multiple this paper, novel Multi-Scale Residual Deep CNN (MSRD-CNN) designed learn manipulation features adaptively for detection. Our network comprises three stages: pre-processing, hierarchical high-level feature extraction, and classification. Firstly, multi-scale residual module employed pre-processing stage extract prediction error noise adaptively. Afterwards, obtained processed by extraction having Feature Extraction Blocks (FEBs) tampering features. Lastly, resultant map provided fully-connected dense layer experiment results show our model surpasses existing schemes even under anti-forensic attacks, when evaluated on large-scale datasets considering proposed provides overall classification accuracies 97.07% 97.48% BOSSBase Dresden datasets, respectively.