I was recently interviewed by Ariel Bogle working for the ABC on an article relating to my PhD Thesis on Photo Response Non-Uniformity noise.
To understand how this fingerprint works, you must go inside the device, into the camera’s sensors.
“The image sensor is broken up into little photosites,” explained Richard Matthews, a PhD candidate researching digital image forensics at the University of Adelaide.
A photosite is a cell in the sensor that detects light, and a camera sensor can have millions of them.
“Each one of those photosites ideally would react in exactly the same way. But … there are slight variances,” he added.
Because of sensor imperfections, a photosite in the top left-hand corner of the image sensor, for example, would not necessarily react in the same way as one in the bottom right.
A PRNU fingerprint can be extracted from a large sample set of images known to be taken with the camera in question.
A filter is then used to create a noise-free image, which is subtracted from the original so all that is left is the noise.
“Then we average a large sample of these noise residues — the left-over noise — and all the other aspects of noise cancel each other out and we’re just left with a PRNU,” Mr Matthews explained.
If it is the same camera, then the PRNU should match the suspect image.
You can read the article in full here where I also talk about how we might use this in conjunction with other pieces of information at our disposal to create solid evidence in a case.
This is a fascinating area of research and I hope to announce some of my research in the area accepted for publication shortly.