Hybrid Teacher Learning Optimization Enabled Deep Learning for Copy-Move Image Multiple Forgery Detection

Abstract

The availability of picture editing software makes it simple to adjust and modify digital photos. A copy-paste forgery, which is used to hide items or create a scene that does not exist, is the most popular method of picture manipulation. There are numerous ways to spot this fake, but it is a laborious and challenging process. To address this problem, multiple forgery detection utilizing copy-move images is discovered and deep learning is enabled through a hybrid optimization technique. The deep learning for copy-move image multiple forgery detection using hybrid teacher learning optimization is presented in this research. Here, the input image is initially taken from Many Images Splicing Dataset, where multiple items are found using the You Only Look Once v3 algorithm to generate an anchor box. Additionally, ZF-Net is used to extract features from each item. Here, ZF-Net parameters are altered utilizing Hybrid Teacher-Learning-Based Optimization (HTLBO). Additionally, the performance of HTLBO_ZF-Net is examined using a variety of evaluation metrics, including accuracy, True Positive Rate, True Negative Rate, Positive Predictive Value and Negative Predictive Value, and it is discovered that these metrics have achieved values of 95.8%, 89.3%, 89.1%, 96.9% and 96.6%, respectively.