Blind forensics tool of falsification for RAW images

This paper presents a novel method for blind forgery detection of natural image in RAW format. The approach is based on a statistical noise model of natural RAW images. This model is characterized by two parameters which are used as a fingerprint to falsification identification. The identification is cast in the framework of the hypothesis testing theory. For practice use, the Generalized Likelihood Ratio Test (GLRT) is presented and its performance is theoretically established in case of unknown parameters where an estimation of those parameters is designed. Experiments with simulated and real images highlight the relevance of the proposed approach.