Automated fingerprint matching can be highly accurate
A study by the National Institute of Standards and Technology showed that computerized fingerprint matching can be highly accurate, outperforming facial-recognition systems.
'With current technology, the most accurate fingerprint systems are far more accurate than the most accurate face recognition systems,' NIST reported in a recent summary of test results
The top performers for fingerprints, with consistent accuracy rates above 98 percent, were systems from NEC Technologies Inc. of Itasca, Ill.; Sagem Morpho Inc. of Tacoma, Wash.; and Cogent Systems Inc. of Alhambra, Calif.
The Fingerprint Vendor Technology Evaluation 2003 was conducted in October and November. The USA Patriot Act requires NIST to certify biometric technologies that could be used in the U.S. Visitor and Immigrant Status Indicator Technology Program. Similar tests on facial-recognition systems were conducted in 2002.
A total of 34 systems submitted by 18 companies were tested, using operational fingerprint data from several federal and state government sources. More than 48,000 sets of prints representing 25,309 individuals were used in the tests. The systems were configured and operated for the tests by company personnel at NIST facilities in Gaithersburg, Md.
The systems were put through a series of tests using single-finger prints, prints from two fingers and sets of prints from four fingers or more. Not surprisingly, NIST found that the number of fingers used and the quality of the prints made a significant difference in the results.
Given a false-positive rate of 0.01 percent, the NEC system had an accuracy rate of 98.6 percent for single-finger prints, 99.6 percent for two-finger prints and 99.9 percent for four, eight or 10-print sets. Both the Sagem and Cogent systems had accuracy rates higher than 95 percent.
In contrast, the most accurate facial-recognition systems tested in 2002 had accuracy rates of 71.5 percent with a false-positive rate of 0.01 percent. To reach an accurate matching rate of 90 percent, the facial systems had a false-positive rate of 1 percent.
But not all fingerprint systems are created equal. NIST reported a large performance gap between the best and the rest.
'Accuracies below 80 percent were typical among the lower third (by rank) of participating systems,' the study found.
Although quality of prints made a difference in accuracy of the matches, the type of scanner or acquisition device used to gather the prints did not necessarily determine the quality of the resulting print. The tests did not evaluate the reliability of scanners or other acquisition devices and did not measure cost effectiveness or address latent fingerprint identification.