Deep Convolutional Neural Network with ResNet-50 Learning algorithm for Copy-Move Forgery Detection

One of the most well-known kind of image forgery is copy-move forgery (CMF). In CMF, one segment of an image is copied and pasted into another segment of an identical image in order to hide crucial data or include new data in the image. For that reason, falsification revealing methods are necessary to recognize altered territories. A new, quick, and capable copy-move forgery detection (CMFD) method is constructed and tested in this study. To classify CMF in a rapid and reliable manner, the proposed method utilizes a Deep Convolutional Neural Network (DCNN) that integrates Residual network framework with 50 layers (ResNet-50). The residual network uses the concept of skip connection which adds the original input to the output of the convolutional layer and will remove the problem of vanishing and exploding gradient that occurs in the traditional Convolutional Neural Network (CNN) model. The accuracy and logarithmic loss (LogLoss), which is a function of cross-entropy, have been employed to assess the performance and efficiency of the algorithm. The results are evaluated on the CoMoFoD dataset after dividing it into training and test set. According to the results, the proposed algorithm achieved better accuracy and reduced the LogLoss function at 100 epochs.