When an Illinois driver gets his or her photo taken at a Department of Motor Vehicles (DMV), the picture is sent to an enormous database and compared to other photos to find potential identity fraud.

The Illinois DMV's biometric-based facial recognition technology is part of a growing trend that has spread to approximately 20 states. Biometrics is the study of methods for recognizing individuals based on one or more physical or behavioral traits.

Illinois has been a leader in facial recognition programs at DMVs. It was among the first in the country to adopt such a program and has become a model for other states. The Illinois program has been a huge success, finding more than 5,000 cases of identity fraud from 1997 to 2007. Of the cases, 4,600 involved people with one fraudulent identification, and 600 cases involved people with two or more fraudulent identifications.

"We think we're getting those licenses off the streets, finding fraud that may not otherwise be found, addressing traffic-safety issues and hopefully reducing overall crime," said Beth Langen, administrator of policy and programs in the Driver Services Department of the Illinois Secretary of State's Office.

Immediate Success
The program started in 1997, when the Illinois secretary of state took measures to detect and prevent identity theft statewide. When the agency decided to revamp the state's DMV services, the resulting RFP required a program to track identity fraud. The department first considered using fingerprint readers before it opted for facial recognition technology because pictures are a generally accepted part of visiting a DMV.

The Illinois Driver Services Department and the Secretary of State Police Department built the facial recognition program from the ground up with technology from Viisage, now part of L-1 Identity Solutions, which continues to maintain Illinois' program. The vendor helped build a database of facial recognition information from licenses that were either new or renewed with the department. It then implemented FaceExplorer, a face recognition application designed to handle large database mining and image management. The application looks for duplicate images from drivers' licenses and identification cards.

"It was an immediate success in that we immediately found instances on our database where people had, in fact, come in and gotten a driver's license and identification under different identities," Langen said.

When an Illinois DMV office captures an image, it's sent directly to the central image server, where the software creates a template for the picture it matches with thousands of other images in the database. The analysis is conducted within 25 servers by using algorithmic facial recognition, first matching images on previous licenses and identification cards, then matching the photos to other Illinois drivers' licenses with different names. Matches are sent to the Illinois Secretary of State Driver Services Fraud Unit, which conducts further analysis to determine whether there is a case for identity fraud.

The solution also enabled the secretary of state and law enforcement investigators to compare facial images from the driver's license image database with video surveillance photos, snapshots, sketches and composite photos.

The Illinois facial recognition program has helped detect thousands of fraud cases, including financial crimes, and on occasion, auto-theft rings, gang activity and welfare fraud. Oftentimes, fraudsters use multiple identities to commit crimes and escape detection.

In one instance, the program helped identify an Illinois resident who had seven false identities and purchased high-end automobiles to operate an auto-theft ring. The person was caught and received a 10-year prison sentence.

In 2005, the program's facial recognition matching dramatically improved when new cameras were installed and the L-1 Identity Solutions software was upgraded. The old version of the system identified three to 20 matches, whereas the upgraded system has a more refined matching capability and rarely displays a wrong match, Langen said. The new system is also speedier: It runs