Fingerprinting Under Fire
Last autumn, Stanford professor Lawrence Wein detected a serious problem in the federal government’s US-VISIT program, designed to capture terrorists entering American airports by checking their fingerprints. While the system is 96 percent accurate overall, Wein has found that its performance degrades when fingerprint quality is not good.
Wein, the Paul E. Holden Professor of Management Science, presented his findings to the White House and the Department of Homeland Security and testified before two subcommittees of the U.S. House of Representatives Select Committee on Homeland Security.
Using mathematical models, Wein, along with Manas Baveja, a doctoral student at the Institute for Computational and Mathematical Engineering and a science fellow at the Center for International Security and Cooperation, has specifically determined that when image quality is poor, accuracy drops to 53 percent, according to an article in The Stanford Report.
“About 5 percent of the general public and 10 percent of those on the watch list have bad quality fingerprints due either to genetics or hard labor,” Wein says. “We assume that terrorist organizations will eventually defeat the US-VISIT program by employing a majority of people whose fingerprint quality is either naturally bad or deliberately made so.”
Wein and Baveja developed various mathematical models that calculated how the system could be tweaked to improve accuracy while not increasing either visitor waiting times at airports or the need for more customs staffing. “We found that instead of scanning two index fingers, scanning 8-10 fingers will result in a 95 percent detection probability, even when fingerprint quality is bad,” Wein says.
Increasing the staffing level of inspectors offers only minor increases in the detection probability for these two strategies, he adds. “Using more than two fingers to match visitors with poor image quality allows a detection probability of 0.949 under current staffing levels, but may require major changes to the current U.S. biometric program,” Wein says.
In the meantime, Wein has proposed a shorter-term solution that will require only a minor software modification. “By loosening the detection thresholds on poor images you can catch more of these people,” he says. “You make up for the additional secondary inspection time this takes by slightly raising detection thresholds on good images.”
A professor at the Graduate School of Business, Wein details his research in an article titled “Using Fingerprint Image Quality to Improve the Identification Performance of the U.S. Visitor and Immigrant Status Indicator Technology Program,” co-authored with Manas Baveja. The paper appears in the May 9-13 issue of the Proceedings of the National Academy of Sciences and can be found at www.pnas.org.