A Generalized Model for Identifying Fake Digital Images through the Application of Deep Learning

Digital photography usage plays a major role in today’s daily communication. Digital picture modification and tampering have become a lot easier because of enhanced technology in photo editing. The public’s perception, political views, and police investigations are all influenced by the manipulation of photographs. Forgery or tampering of digital photograph can be done by numerous techniques such as image splicing, copy-move, JPEG compression, light inconsistencies, retouching, etc. Every manipulation technique will leave some traces after forgery and with the help of those traces detection of forgery can be done by forensics. Digital forensics faces difficulty in identifying the authenticity of such digital images. Though manipulation of a single image is done by applying numerous techniques, traces are erased by using post-processing. Many approaches are proposed for image forgery detection by using any one of the tampering techniques (like copy-move, splicing, etc.). In the real–world scenario a universal model is recommended for detecting multiple manipulation operations on a single image.This paper aims to help detect forgery more effectively in digital images even with multiple tampering techniques used to forge one single image using deep learning techniques. Deep learning techniques take a large set of original and tampered images for forgery detection. This model is a self-learned model which differentiates the two classes of authentic and forgery images by extracting hidden features. The experiment results of the proposed approach will illustrate the model’s accuracy and efficacy.