Decomposed PRNU Library for Forensics on Photos

Abstract

Today, the digital forensic techniques for digital images are developed with the origin identification and integrity verification functions for security reasons. Methods based on photo-response-non-uniform PRNU are widely studied and proved to be effective to serve the forensic purposes. However, due to the interpolation noise, caused by the colour filtering and interpolation function the accuracy of the PRNU-based forensic method has been degraded. Meanwhile, the tremendous physical storage requirement and computation consumption limit the applications of PRNU-based method. Therefore, an innovative DPRNU-based forensic method has been proposed in order to solve the above problems. In the method, the artificial component and physical component are separated according to the colour filtering array CFA and the PRNU are only extracted from the physical component in order to remove the interference caused by the interpolation noise, which increases the accuracy of the camera identification and integrity verification. Meanwhile, due to the separation, the DPRNU are only 1/3 of the size of the traditional PRNU, which saves considerable physical storage in setting up the digital library and fasters the comparison speed between the fingerprints.