Inpainting and forgery detection on images using deep learning techniques

The need to identify altered or phony photos is greater than ever due to the increasing strength and accessibility of image editing and alteration tools, particularly in domains such as digital forensics, media verification, and image repair. This study explores a method that brings together two different techniques: one that modifies parts of an image (using a process known as Telea- based inpainting), and another that identifies forgeries through a machine learning approach called DCGAN. The inpainting tool helps simulate how real image tampering might happen by reconstructing damaged or manipulated areas in a realistic way. Next, the detection system is trained to recognize small inconsistencies in texture and structure, learning to tell the difference between real and fake by comparing them. After training on a dataset of 6,000 images over 50 learning cycles, the system reached an accuracy rate of 95.52%, showing strong potential for catching image forgeries automatically and accurately.