Improving the reliability of digital camera identification by optimizing the algorithm for comparing noise signatures.

In this paper, the issues on current optimization methods for identifying modern digital camera photosensors are presented. Methods for improving the reliability of identification of digital cameras are discussed, i.e., determining the digital camera used to take a particular photograph. Homogeneous images were optically recorded to form a noise signature, and sets of amateur images were tested for identification of three types of cameras from the images. Digital image filtering and image identity metric were optimized. An optimal digital filter was selected to evaluate smoothed images to obtain the noise signatures of the identified cameras. An optimal identity criterion was obtained by comparison of camera noise signatures. Using an optimal filter and identity criterion allowed increasing, on average, the reliability of identifying the camera used to take a particular photograph by > 60 times. This work provides reference to solve image problems in image registration and processing, security, forensics, and big data analysis. [ABSTRACT FROM AUTHOR]