NIST releases results of latest fingerprint-matching tests
- By William Jackson
- Feb 19, 2008
Only one of five biometric smart-card vendors participating in a federal test of on-card fingerprint matching managed to meet the accuracy standards set for the federal Personal Identity Verification (PIV) card, according to the National Institute of Standards and Technology.
A card from Sagem Morpho Inc., of Tacoma, Wash., was the best-performing product in the results
NIST released Monday.
NIST conducted the second phase of the Minutiae Interoperability Exchange Test (Minex) last year to evaluate the performance of on-card fingerprint verification. The first phase
of testing was done in 2004 to test off-card verification. The Homeland Security Department's Science and Technology Directorate sponsored the tests.
They evaluated the speed and accuracy of fingerprint-matching algorithms that comply with International Organization for Standardization standards. The cards are used to control physical and logical access at government agencies. Homeland Security Presidential Directive 12 requires that agencies issue PIV cards containing fingerprint data to government employees and contractors. NIST has established technical standards for the cards.
The Minex tests were benchmarks and did not include interface standards, secure transmission protocols, vulnerabilities or other variables. The cards used matching algorithms supplied by five companies. They were tested with fingerprint templates containing the minutiae ' or defining traits of the fingerprints ' created by a variety of companies. Each test involved two templates, one with the print being submitted for verification and one with the print it is matched against.
In the tests, accuracy was higher when both fingerprint templates came from the same vendor, which might have given Sagem Morpho an edge. Its card was the only one that used its own matching algorithm and had its own templates in the test.
One other card ' from Oberthur Card Systems with a matching algorithm from id3 Semiconductors ' came close to meeting the PIV standards.
NIST tests indicate that fingerprint matching done off-card could be more accurate than when it is performed on-card. Only Sagem Morpho submitted cards in both phases of the Minex tests. The report states that 'match-off-card error rates are all lower than match-on-card error rates,' although the differences might not be statistically significant.
Accuracy is a factor of two elements: the rate of false acceptances, or false positives, in which a submitted fingerprint is improperly found to be a match; and false rejections, or false negatives. In comparing the Sagem Morpho results for the 2004 tests and the more recent ones, the match-on-card algorithms produced 20 percent to 40 percent more false negatives than the match-off-card algorithms.
When sensitivity of the matching process was set at its highest ' to produce only one false positive in 10,000 tries (0.0001) ' the false negative rate was two to four times greater than when the false positive rate was set at 0.01. But cards from the same vendor behaved comparably at both settings. A card that performed better at a false positive rate of 0.0001 also performed better at 0.01.
NIST plans to conduct a third phase of Minex testing later this year.
William Jackson is a Maryland-based freelance writer.