PhotoDNA, a new child-pornography tracking technology developed in part by Dartmouth computer science professor Hany Farid, may soon enable Internet service providers to better curb the availability of online images depicting child pornography. The technology was donated to the National Center for Missing and Exploited Children in December.
Farid, who has been working with Microsoft for five years on digital forensics, began his work on PhotoDNA after Microsoft decided to "make an effort to remove the content that is illegal or inappropriate" from the Internet, Farid said. The technology will be implemented by several Internet service providers over the next few months, he said.
"Any time somebody tries to address a problem on the scale of the Internet it can be very intimidating because of the sheer magnitude," Farid said. "I thought it was technically interesting, I thought it was important and I thought it was something we could solve."
PhotoDNA will track images of prepubescent children engaging in "increasingly violent and sexual" acts, according to Farid. The technology, however, will not be used to flag photographs of 16- or 17-year-old individuals participating in sexual acts.
The technology and core algorithms were first created by Microsoft Research. Farid's involvement entailed "large scale testing, refinement and making the code run faster," he said, explaining that the PhotoDNA technology must be efficient since it will track billions of photos.
"If we tried to solve this problem by creating a technology that could mathematically look at a picture [and determine] that this is a child in an inappropriate act that would be nearly impossible," Farid said.
The task is difficult because computers cannot distinguish between photos a parent posts of his or her child taking a bath and those of child pornography, he said.
PhotoDNA will not search for new photographs of child pornography but rather will enable Internet service providers to identify and remove images that are already known to contain child pornography. Many of the images, which have already been flagged by NCMEC, are still available online because they have been altered or cropped, Farid said.
PhotoDNA employs a unique mathematical signature known as an image's "DNA" for each photograph using technology that is more advanced than previous tracking methods. Each image will be labeled with a "DNA" that is distinct to that photograph and will remain the same despite minor changes to the photographs through cropping or other forms of digital editing, Farid said.
Although PhotoDNA is able to track billions of photographs at a time, it will not replace human involvement in the tracking of child pornography available online, Farid said.
"There is always a human in the loop who looks at the image and makes an assessment and then makes a decision," he said.
PhotoDNA may also be used by law enforcement agencies to track child predators.
"NCMEC is equipped to make the PhotoDNA tool available to law enforcement agencies, online service providers and others working with NCMEC to disrupt the ability of predators to use the Internet to exploit children or traffic in child pornography," NCMEC president and CEO Ernie Allen said in a press release.
Farid said that in the future he hopes to develop a similar technology that will allow Internet service providers to track videos of child pornography that are available on the Internet.
Allen could not be reached for comment.
