Digital Camera Identification by Modeling the Probability Distribution of Sensor Pattern Noise

Sensor pattern noise (SPN) as the fingerprint of imaging devices, could be used as a reliable feature in digital source identification. In this paper, we introduce a new method which uses the probability model of SPN to identify the source. To achieve this goal, after extracting the SPN of some images, they are digitized and then for each value, the distribution of its neighbors is modeled separately ("Value-Model"). Finally by using the value-model, the quantized SPN is mapped to the probability domain. The average of probability matrix of some images of same camera forms camera model. This SPN model causes noticeable increases in the true detection rate of the source. To evaluate the efficiency of proposed approach, we do some benchmark on our hypothesis. The accuracy and performance of our model compared to similar works, proves high efficiency of the proposed theory.