When a gun is used in a crime, forensic investigators identify it by the unique pattern of scratches that its barrel leaves on bullets. A similar trick is now being used to match digital images to the cameras that captured them, an important advance as child pornography crimes increase.

Software developed by Jessica Fridrich at the State University of New York in Binghamton exploits the fact that every digital camera introduces a unique pattern of imperfections, or “noise,” into its images. In monochrome areas, for example, individual pixels might actually be slightly different colors. Fridrich’s software determines a camera’s noise signature by identifying the irregularities in its pictures. That yields a “fingerprint” that investigators can search for in other photos.

Fridrich tested her software using 10 cameras and a total of 3,000 pictures. In every case, the software matched the picture with the right camera, she says. “This is very nice work in the exciting and important problem of camera ballistics,” says Hany Farid, computer science professor at Dartmouth College.

Fridrich is currently seeking a patent and says the FBI is evaluating the technology as an investigative tool. The need is great: between 1996 and 2002, the number of federal cases involving child pornography exploded from 113 to 2,370, and the FBI predicts the trend will continue.