Image Source Identification Using Machine Learning Techniques

Abstract

In digital forensics, identifying the model of the camera that was used to obtain a given digital image, presents an interesting problem. In a scene of a crime, it may help the forensic investigator to identify the camera that has taken the digital image used as evidence. The properties of the image given in the Exchangeable Image File Format are easy to modify and thus, cannot be authenticated. The exponential increase in digital image capturing devices in the past few years has further added to the problem. In this work, we aspire to provide experimental results and compare the accuracies generated by a number of two-class classifiers when fed with multiple features of images captured by two camera models, which help in the identification of the source camera model of the given image. Hence, to improve the efficiency of source identification, we strive to improve upon the classifiers in use.