Comparison of Deep Learning Networks for Source Camera Identification

Digital Image Forensics is getting more attention in the research area as the need for confirming the authenticity of images is increasing with the increased use of information sharing through social media. Source Camera Identification (SCI) is an important Digital Image Forensic technique to ensure the identity of the source/camera which has captured a particular image. SCI methods are hot topics in research area for two decades, but with the advancement of Convolutional Neural Networks for Computer Vision applications, so many deep learning based techniques for Source Camera Identification is proposed. This paper aims to review the important deep learning networks for Source Camera Identification. In this paper, we describe the basic structure of the algorithm for Source Camera Identification and present a detailed comparison of various deep learning based techniques on the basis of accuracy, number of cameras and dataset used. Also a source camera device identification technique is implemented using VGG16 model witha an accuracy of 87.16% which is higher than the state of the art methods.