Sensor pattern noise based source anonymization

Abstract

In today's digital world, images are utilized as a method of communication in all spheres of life. Counter Forensics is the art and science of impeding and misleading forensic analysis of digital images. Camera sensor pattern noise is efficient in blind image source device identification. In this paper, we deliver an attack on digital images, where we completely remove the traces of sensor pattern noises of their source devices, so as to deceive forensic investigations. Next, we substitute the sensor pattern of a given image with that of a different (wrong) source device, such that it now appears to the forensic analyst, that the image was produced by device B, whereas originally it was produced by A. Our experimental results prove that high correlation is achieved between a tampered image and a wrong device, suggesting considerably high degree of anonymity, hence misleading forensics investigation.