iStockPhoto

Czech police nab a man suspected of raping 12-year-old girls after offering them car rides via an Internet Web site. In Ohio, a 400-pound man, likewise, uses a Web site to impersonate a 15-year-old boy in order to convince a 12-year-old girl to send photographs and videotapes of herself naked. A sting operation in the U.K. shuts down a pedophile chat room Web site, and the site's leader is caught with over 75,000 pornographic images.

Social networking over the Web has helped connect millions of Internet users, but all of this online interaction can also have a serious downside: a proliferation of pedophiles who use code words to trade in child pornography or prowl chat rooms and befriend underage victims, peppering their messages with words like "kewl" and other youthful colloquialisms.

In a move that pits technology against criminals (and, some fear, privacy), a group of researchers at Lancaster University in England and law-enforcement officials at the United Kingdom's Child Exploitation and Online Protection Center  (CEOP) is developing software that tracks the Web's evolving child pornography lexicon as well as predators' chat strategies to help law-enforcement agencies catch the most secretive of these criminals before they strike.

"There's a list of about 50 key words that are very indicative of child pornography," says Doug Skinner, a forensics expert who works at Immigration and Customs Enforcement’s Cyber Crimes Center in Fairfax, Va.,* and sometimes coordinates with CEOP as part of the Global Virtual Task Force. But, he says, "The terms do change."

That's why Awais Rashid, a Lancaster University computer science professor, has launched the three-year Isis Project, which uses linguistic analysis to keep tabs on these Internet-savvy pedophiles. "There's so much activity it's virtually impossible to police," he says. Currently, investigators at ICE (U.S. Department of Homeland Security) and CEOP are left waiting for a potential victim to report suspicious activity, but by then, it is often too late.

Rashid's strategy is to create automated monitoring tools that operators of chat rooms, social networks, and file-sharing networks would install on their sites. This will provide law-enforcement officials with an automatically updating dictionary of these code words, along with an alert system which will inform them when users are detected masquerading as children.

In preparation to write the software module for file-sharing networks (a prototype by the end of the year), the team sifted through an entire month of search traffic on the number one peer-to-peer file-sharing network, Gnutella, between February 27th and March 27th 2005. Because each peer participates in routing network messages to and from other peers, Rashid's team could set up a specialized client to intercept and log these queries throughout large segments of the network. Then, two specialists at CEOP analyzed 10,000 keyword searches from three separate days to determine whether they contained references to child pornography. About one in every 100 searches was for such material, and about 1.6 percent of search results received contained such material. Because of the size of the Gnutella network, which had a population of 1.81 million users that year, thousands of child-pornography related searches are being conducted every minute. For comparison, ICE arrests about 2,500 child predators in the U.S. per year.