Binghamton researcher uses photos to link cameras to digital crime
The process of matching bullet to gun is exhaustively detailed over the entire 1967 episode of Dragnet, "The Gun." LAPD Detective Joe Friday, investigating the murder of one Keiko Hashimoto, has two bullets: one recovered from the body, the other from a gun found in the house of a suspect.
The verdict, delivered in the Dragnet characters’ classic, early crime drama cadence:
“Tuesday, August 30. 6pm. Those bullets came from this weapon. The striations match perfectly.”
Today’s the process is such a standard of TV crime drama that all a fictional forensics team-member needs to say is "the bullet’s a match."
But for the past decade researchers and investigators have been using digital methods to "find a match," linking photos to the cameras that committed the crime.
"Imperceptible to the human eye"
“We try to discover things that are imperceptible to the human eye ... the 'inaudible' in digital images,” explains Jessica Fridrich, a researcher at Binghamton University.
Fridrich entered the field of digital forensics early in her career, and her research team in Binghamton is continually at work on the mathematical models used to analyze photographs to link them to the cameras that took them.
Fridrich brings up a picture on her computer screen to illustrate what her work produces. A block of random colored squares looks like a 1980s vintage screensaver. But it's actually a visual representation of the tiny variations within a camera’s imaging sensor.
“There are very small differences, but nevertheless [they're there]," Fridrich explains. "For example, one pixel may consistently register 2 percent larger than neighboring pixel because it’s 2 percent larger. And this irregularity is consistent from picture to picture.”
Camera-makers actually know these imperfections exist. They’re just so small the companies don’t bother removing them. This comes in handy for Fridrich: all it takes is a little computer analysis, applied to irregularities gleaned from a series of photos, and presto, the digital "fingerprint" of a camera.
“If you end up catching the pirate in a movie theater, you can actually associate the specific video camera with other pirated content that was done previously,” she says.
Child pornography is another major area where tracking down digital fingerprints has been useful. Fridrich’s research team was recently brought in as an expert witness in a take-down of a major European ring.
Mark Pollitt, who spent twenty years as an FBI special agent and now teaches at Daytona State, says camera fingerprinting can be applied to crimes running the gamut, from kidnappings to industrial espionage.
He says when digital cameras were first introduced, they were an expensive proposition for everyone but the biggest photography nerd. But now prices have dropped dramatically, and digital cameras are ubiquitous.
“I’ve got an iPad and I can’t figure out why I would want a camera in it," Pollitt says. "[But] everyone else who buys one from here on out is getting one whether they want it or not. And so what’ll eventually happen is people will start to use that."
Meaning, inevitably, the more photos we take, the more we leave traces of our lives in digital form, for researchers to pick apart.