<p>A broad range of applications have recently drawn the interest of researchers studying digital image processing. Image falsification is one famous area where experts are concentrating their research. They concentrate on copy-move image forgery that involves deception. Copy-move image forging is the method of creating a fake image by copying a component of the original image and inserting it into the fake image. This research proposes a metaheuristic-assisted Convolutional Neural Network (CNN) model to detect the Copy-Move forgery. The following approach consists of three steps: copy-move forgery detection, feature extraction, and image augmentation. Subsequently, the features extracted from the augmented image include Speeded-Up Robust Features (SURF), Discrete Wavelet Transform (DWT), and proposed Local Vector Pattern (PLVP) features. According to the proposed LVP, an adaptive CST is introduced to enhance its adaptability to local variations. An adaptive CST is added in accordance with the suggested LVP to improve its flexibility in response to regional differences. Lastly, an optimized CNN model is trained using the retrieved features to detect copy-move forgeries. By adjusting the ideal weights, the CNN model is trained using the Improved Moth Flame Optimization (IMFO) algorithm. For Normal and Forgery images, the detection model's final output is labeled 0 and 1, respectively. The IMFO approach's FDR of 0.035 is notably low when looking at the training data at 80%. In contrast, the conventional methods, Pelican Optimization Algorithm (POA), Coati Optimization Algorithm (COA), Osprey Optimization Algorithm (OOA), Tasmanian Devil Optimization (TDO), and Improved Moth Flame Optimization (MFO) have shown relatively higher FDR ratings of 0.126, 0.139, 0.132, 0.157, and 0.112, respectively.</p>