Autoregressive-Elephant Herding Optimization based Generative Adversarial Network for copy-move forgery detection with Interval type-2 fuzzy clustering

Abstract

A copy move forgery is considered as the general form of digital image forgery, where a portion of the image is copied as well as pasted in some other position in the same image. To act upon forgery is simple; however detecting the forgery is more complex due to the copied portions’ features similar to other parts of images. Therefore, an effectual forgery object detection approach is exploited by exploiting the adopted Autoregressive Elephant Herding Optimization based Generative Adversarial Network (A-EHO based GAN). The proposed A-EHO approach is derived by incorporating Conditional Autoregressive Value at Risk by Regression Quantiles (CAViaR) with Elephant Herding Optimization (EHO). First, features like Local optimal oriented pattern (LOOP) and Convolutional Neural Network (CNN) features are extracted for each foreground object. Then, the features are employed to the GAN for the computation of the forgery score. Here, the RideNN classifier detects the forgery image based on the feature vector and the forgery score. As a result, the adopted approach achieved higher performance regarding detection rate, ROC, TNR, as well as TPR with the values of 96.687%, 98.35%, 97.809%, and 96.971%, respectively.