Deep Learning to Detect Image Forgery Based on Image Classification
Journal of System and Management Sciences · 2022
Nowadays, Digital images can be seen in magazines, newspapers, hospitals, shopping malls, on the Internet, and among other places.As technology advances, at the same time, the trust in images is decreasing day by day because the easy to forgery in these images.One of the major topics for the researcher is the detection of forgery in images, and copy-move (CMFD) is one of main types of image forgery.The majority of CMFD algorithms now in use relies on keypoint or block approaches, individually or merges of them.Many deep convolutional neural network (CNN) methods have recently been used in image classification and image forensics to outperform more conventional techniques.In this paper, we proposed a new method for image forgery detection using a CNN.CASIAV1, CASIAV2, and Columbia datasets are used in the proposed method.The pre-trained (CNN) is used to extract dense features from the test images for Support vector machine (SVM) and K-Nearest neighbor (KNN) classification.The model is designed, implemented and tested.From the experimental results we can observe that the accuracy is 98.22% for CASIAV1, 97.02% for CASIAV2, and 85.1% for the Columbia dataset.