An Optimized Fuzzy C-Means with Deep Neural Network for Image Copy-Move Forgery Detection

Abstract

Copy Move Forgery Detection (CMFD) is one of the significant forgery attacks in which a region of the same image is copied and pasted to develop a forged image. Initially, the input digital images are preprocessed. Here the contrast of input image is enhanced. After preprocessing, Optimized Fuzzy C-means (OFCM) clustering is used to group the images into several clusters. Here the traditional FCM centroid selection is optimized by means of Salp Swarm Algorithm (SSA). The main inspiration of SSA is the swarming behavior of salps when navigating and foraging in oceans. Based on that algorithm, optimal centroid is selected for grouping images. Next, the unique features are extracted from each cluster. Due to the robust performance, the existing approach uses the SIFT-based framework for detecting CMFD. However, for some CMFD images, these approaches cannot produce satisfactory detection results. In order to solve this problem, the current method utilizes the stationary wavelet transform (SWT). After extracting the features, the CMFD detection is done by RB (Radial Basis) based neural network. Additionally, it is computed by means of diverse presentation metrics like sensitivity, specificity, accuracy; Positive Predictive Value (PPV), Negative Predictive Value (NPV), False Positive Rate (FPR), False Negative Rate (FNR) and False Discovery Rate (FDR). The proposed copy move forgery detection method is implemented in the working platform of MATLAB.