From Criminals to Web Crawlers

A search-engine technology developed with the help of the FBI has made a dramatic impact on police investigations. Now it's set to take on what could be a more daunting challenge: the Web. By Kristen Philipkoski.

A crime-fighting search engine used to fight terrorism and insurance scams may soon find a home at one of the Web's top search engines. The system, called VCLAS, has helped detectives crack cases all over the world.

"In 11 days, the PhoneFraud software helped law-enforcement agencies in New York uncover US$1.2 billion in stolen services," said Jay Valentine, president and CEO of InfoGlide, the company that owns the VCLAS software package.

The software is built around a "Similarity Search Engine," which thrives on imperfect and complex information, data that engineer David Wheeler said often stumps search algorithms based on neural networks.

Similarity searching is well-suited to crime work, Wheeler said, because investigations are often inherently random and disconnected. For instance, if police are looking for a red vehicle, but a witness says it was maroon, a traditional keyword search wouldn�t register a match since it couldn't recognize that the colors are similar.

With some help from the FBI, Wheeler invented the first version of the system -- called Detective Toolkit -- in the wake of his father's 1989 murder. He began cutting his own code when he discovered the then-limited search tools that detectives had at their disposal.

Since then, the program has helped solve countless serial crimes, as well as insurance fraud cases. In April 1997, InfoGlide representatives worked pro bono to help bust a huge wireless telecom fraud operation. The US government uses the software to investigate terrorist groups.

A corresponding system used by the FBI -- called VICAP -- has identified four crime series (all the crimes a serial criminal commits). VCLAS has identified well over 1,100, said Wheeler.

The system obviously works for cops, but Valentine thinks the technology has a market beyond criminal investigations.

InfoGlide is in talks with one of the top six Internet search companies to license the company�s forthcoming Internet search engine, Tornado, which is under development, Valentine said.

The new program will be coded in the Java language, and Valentine hopes to ship a beta release by the beginning of next year. The Internet application will work in a similar fashion to the way InfoGlide's database technology works to solve crimes.

A Tornado end user would create "profiles," which might contain one, two, or hundreds of variables to identify a subject, person, or other type of target. Unlike current search systems, where the user chooses five or six topics of interest, he or she will tailor Tornado profiles to meet specific needs, Valentine said.

"It becomes a live robot on your PC, in real time," Valentine said.

But one search engine expert said that the company may have a tough challenge ahead.

Danny Sullivan, editor of the industry newsletter Search Engine Watch, says the technology sounds similar to that owned by Firefly and Autonomy. Both allow users to personalize searches and build profiles.

"This is interesting and useful for many people, but I doubt it will revolutionize Internet searching," said Sullivan. InfoGlide may find that analyzing data amid the scattered, constantly updated pages on the Internet is an even more uncertain science than navigating intricate crime-scene database records.

"Unless [Valentine has] indexed 60 million Web pages he doesn�t know how things happen in the real world," said Sullivan. "It�s useful if you�re doing research, but the majority of searches are under two words, so they�re not very complicated."

Note: Wired News is owned by Wired Digital, which also owns the HotBot search engine.