In recent years, the development of fake or forged digital images is common by applying different popular image editing tools such as adobe photoshop, affinity photoshop, CyberLink photo director 365, and by using high-resolution capturing devices. It is difficult to differentiate between the original and tampered or fake images. Copying and relocating one or more sections of an image within the same image is referred to as copy-move forgery, which has become a prevalent form of forgery in contemporary times. Conventional techniques for detecting copy-move forgery (CMFD) can be broadly classified into two categories. One is using block-based methods and the other is key point-based methods. These methods have their limitations such as the computational cost being high because of the large amounts of data, and the high error rate in smooth areas of the image. Further these are less robust against pre-processing or post-processing operations and scaling attacks. A lot of research has been done and many techniques exist to localize the forged image and to detect the forged region of images. In this paper, the survey is made on the recent developments in copy move forgery detection techniques using deep learning. Additionally, we explore different benchmark datasets that have been employed by various techniques for the detection of copy-move forgery. We also discuss key aspects and comparison of various deep-learning techniques which are used to detect copy-move image forgery.

Convolutional Neural Networks, Image forgery, Copy move forgery, Deep learning