Abstract
Protecting data against tampering is a significant concern in modern times. Digital photographs are essential for displaying information. Digital picture forgeries include adding unusual patterns to real pictures, causing visual heterogeneity. CMF is a sort of digital image forgery in which a segment of an image is linked to a similar picture to cover or recreate forgeried components. The forgery seems authentic since the goal area has the same qualities as the original. Despite several ways to detect CMFD, there exist research gaps such as false detection, excessive execution time and low accuracy. Therefore, to address this issue, we present a hybrid optimization technique and classifier for CMFD. Proposed a novel deep learning technique stacked sparse denoising autoencoder (SSDAE) to classify the images as legitimate or fake. Additionally, the weight and bias parameters of the SSDAE model are optimized using the Grasshopper Optimization Algorithm (GOA) and the Spotted Hyena optimizer (SHO). The experiments are conducted on MICC-F220, MICC-F600, MICC-F2000 and CASIA2.0 datasets. Experimental results indicate that the proposed scheme find out image forgery region with Accuracy = 97.45%; Precision = 98.75%; Recall = 98.25% and F1 = 98.55% on MICC-F200 dataset, Accuracy = 98.92%; Precision = 88.45%; Recall = 85.21% and F1 = 91.41% on MICC-F600 dataset, Accuracy = 99.12%; Precision = 99.25%; Recall = 91.14% and F1 = 85.32% on MICC-F2000 dataset and Accuracy = 98.02%; Precision = 96.03%; Recall = 97.74% and F1 = 97.48% on CASIA 2.0 dataset.