Image Copy Move Forgery Detection Using Multi-Plane Convolutional Neural Network

In digital images, the digital device usage and availability of open source for image editing applications are easily manipulating the digital images. The Copy Move Forgery (CMF) is an extensive technique for duplicate or hide particular image portion without leaving visual signs. Therefore, it is complex to detect the CMF and the forensic experts are rely on efficient technique for CMF detection. The Multi Plane Convolutional Neural Network (MP-CNN) is proposed in this research for detecting CMF in images. The MICC-F220, MICC-F2000 and CASIA 2.0 datasets are used which is publicly available Kaggle dataset. This dataset is preprocessed by image denoising which reduces the noise from input image. Then, the pre-processed images are provided to MP-CNN for detecting the CMF. The metrics such as precision, accuracy, f1score and recall are applied for estimating the MP-CNN performance. The MP-CNN attains accuracy of 99.45%, 99.26% and 98.61% for MICC-F220, MICC-F2000 and CASIA 2.0 datasets when compared to existing techniques like CNN, Masked Region CNN (M-RCNN) and Contrast Limited Adaptive Histogram Equalization with CNN (CLAHE+CNN).