Not so long ago, those who trafficked in pornographic images of children kept to the shadows, operating their nefarious business far from mainstream channels. Then along came the Internet. The advent of instant publishing and file sharing has opened a global e-marketplace for child porn, with law enforcement lagging far behind tech-savvy traffickers.

Hany Farid, a computer scientist from Dartmouth College, was appalled to learn that not only is this illicit business booming, “but the children are getting younger and the images more violent. This is a problem that technology has gotten us into,” Farid mused. “Let’s see if we can use technology to help get ourselves out of it.”

Farid collaborated with researchers from Microsoft to develop a new tool intended to disrupt online trafficking in child porn. The core technology is called PhotoDNA. It extracts a unique signature from any digital photo using a process called “robust hashing.” This numeric signature, which Farid likens to human DNA, does not change even if a photograph is resized or edited. The signature can be used to identify matches across very large data sets.

The process is automated, meaning no human has to review potentially offensive images. It’s also lightning fast—five milliseconds to extract a signature—and has proven highly reliable in massive testing. For law enforcement and online service providers on guard against child porn, Farid adds, “this means being able to find the proverbial needle in the haystack.”

Microsoft has donated PhotoDNA to the National Center for Missing & Exploited Children (NCMEC), including the right to sublicense the technology to online service providers. In recent years, the nonprofit NCMEC has worked closely with law enforcement to identify nearly 30 million photos of child porn. With the use of PhotoDNA, those images can be used to generate a vast data set of digital signatures to detect known photographs of child pornography.

If online service providers detect any of these images, they can report them to the NCMEC. The long-term goal, according to Sue Hotelling of Microsoft’s Digital Crimes Unit, is to “help stop the distribution of these illegal and horrific images and help stop revictimization of children whose images may otherwise be viewed again and again online.”

Pulling down those images won’t keep new child porn from being uploaded, but it may help to reduce the problem. “People who traffic in child porn seem to pass around the same images, person to person,” Farid says. Getting known images offline “is a little more tractable” than cleansing the entire Internet of child pornography, he says.

Because PhotoDNA is a generic tool, it could be applied to any type of image. “It may have other applications down the road,” Hotelling says. “We are exploring other ways to put it to use, including incorporating the technology into tools to help law enforcement in their child protection investigations,” she adds.

Implementation of the tool is still in the early stages, with Microsoft starting to search public sources for some of the worst known instances of child porn. The goal, Farid says, is to have it implemented “at all the Internet service providers around the world. We’re still working on that.”

Tracker Pixel for Entry