Camera-identification and common-source identification: The correlation values of mismatches

Abstract

A robust and well-known way to identify the source of an image is the use of Photo Response Non-Uniformity. The ability to be able to extract the PRNU-pattern from images is both used for camera-identification (linking one or more images to a camera) and common-source identification (linking images to images). In this paper we focus on the correlation values of mismatches, i.e. the correlation between images made with different cameras. Although the correlation values of mismatches are close to zero, they are never exactly zero. In this paper we show that it is possible with an extremely simple formula to a priori estimate the typical range for mismatch correlation values. This simple formula can be used as a decision rule in digital camera identification to either perform a complete investigation including reference recordings (which is time consuming) or not. In common-source identification this simple formula can be used to provide a well-educated guess for the threshold value to the cluster algorithm instead of just arbitrarily trying a range of threshold values.