New developments in tracing particular photographs to the cameras that snapped them might provide the basis for a forensic method of catching pedophiles who distribute child pornography anonymously on the Internet. It could also help law-enforcement agencies identify smartphone thieves who take pictures with the stolen gadgets and then post the images online.

It has been known since 2006 that tiny variations in the silicon chip–based camera sensors create differences in response to light that leave a signature “noise” pattern (above right) on every photo that can be matched to a specific camera and cannot be removed. “It is not currently possible to perfectly separate the image from the noise, modify the noise and then add it back to the image,” says Riccardo Satta, a scientific officer at the European Commission Joint Research Center's Institute for the Protection and Security of the Citizen. At a recent privacy conference in Brussels, Satta presented work showing that sensor-pattern noise persists when photos are modified and uploaded to social media.

Investigators have long known of other identifiers that digital cameras insert into images as they convert a stream of light into digital bits. But none are as reliable for tracing the source of an image as sensor-pattern noise.

In a preliminary study of 2,896 images taken from 15 different social networks or blog accounts, Satta and his colleague Pasquale Stirparo found that a photograph could be linked half the time to a specific camera as a most probable match. They also discovered that a set of images could be accurately grouped according to the originating camera 90 percent of the time, with a false positive rate of 2 percent.

These statistics are not good enough to use at a trial. But the technique could help select targets for investigation, especially when presented along with other information found on social networks, such as location and friend lists.