Determination of stop-criterion for incremental methods constructing camera sensor fingerprint

Abstract

This paper aims to find the minimum sample size of the camera reference image set that is needed to build a sensor fingerprint of a high performance. Today’s methods for building sensor fingerprints do rely on having a sufficient number of camera reference images. But, there is no clear answer to the question of how many camera reference images are really needed? In this paper, we will analyze and find out how to determine the minimum needed number of reference images to remove the mentioned uncertainty. We will introduce a quantitative measure (a stop-criterion) stating how many photos should be used to create a high-performance sensor fingerprint. This stop-criterion will directly reflect the confidence level that we would like to achieve. By considering that the number of digital images used to construct the camera sensor fingerprint can have a direct impact on performance of the sensor fingerprint, it is apparent that this, so far underestimated, topic is of major importance.