ABSTRACT
The camera model identification is one of the fields addressed in Digital Image Forensics (DIF) that solves a series of forensic issues. Following this cognition, forensics community developed a set of camera model detection algorithm utilizing traces left behind on captured images during the acquisition process. Work for identifying camera model implied by machine learning where features are trained on learning model. However, this can carry out by exploiting deep learning scheme with tradition classification model as Convolutional Neural Network (CNN), which automatically extract features and trained by some learning process. In this paper, CNN based CamMod model is proposed for this identification problem by the use of captured image’s traces. The varied numbers of kernels take each image and extract features automatically. Here model is analyzed taking distinct configuration of parametric layers. Experimental evaluation gives better result for three convolutional layers examined for distinct class groups, moreover the proposed CamMod network compared with existing model providing improved identification accuracy for given 12 and 14 camera model classes.
