Digital pics contain ‘fingerprints’ of cameras used to click them

A digital picture carries pixel-based fingerprint that can reveal which camera model was used to click it, say researchers.

Nasir Memon of the Polytechnic University in Brooklyn, New York, has revealed that every digital camera has an in-built "demosaicing" software that translates each pixel into a usable colour and brightness signal.

He says that software has to be tailored to a particular camera type to cater for the many peculiarities of each model, including colour filter arrays, lenses, and light-sensitive microchip called a charge-coupled device (CCD).

According to him, one of the algorithm''s tasks is to work out the colour a screen pixel should adopt without jarring with the colours of neighbouring pixels.

Memon says that he and his colleagues have discovered how to work backwards from neighbouring pixel values in a photo to identify the model-specific demosaicing algorithm that created it.

The researcher says that his team's approach has been found to identify cameras with 90 per cent accuracy in early tests.

Mark Pollitt, a former FBI crime lab scientist who is currently associated with the University of Central Florida in Orlando, says that the novel technique may make a big difference to detective work if further tests too provide promising results.

He substantiates his statement with the example of a photo of any kidnapped person emailed to a press agency.

"If we can identify the camera, then there is a possibility that we can identify who bought it and where," New Scientist magazine quoted him as saying.

Though the same camera models may be owned by many people, Pollitt still believes that the new approach can be used forensically.

Given that digital cameras have a shelf life of only 18 months, he adds, it may be possible to find out when and where it was sold.

http://www.zeenews.com/sci-tech/technology/2008-11-16/483940news.html