Idemia biometric algorithm generates thousands of leads as fingerprint examiner tests questioned

Michigan State Police say the new fingerprint matching software the department acquired from Idemia in late 2018 is a “game changer,” enabling matches from latent prints that have generated thousands of new leads, Wood TV reports.
“We’ve got a system that just outperforms anything that we’ve been exposed to,” Lt. Rob Hackerd, who supervises the Latent Print Unit at the Michigan State Police Crime Lab, told Wood TV. “Right away we saw that prints that are of a lower quality, we could get in the system and the likelihood of a match was greatly increased.”
An MSP spokesperson said the Idemia system cost nearly $8 million over 5 years, including installation, hosting, and maintenance.
Hackerd compared the difference to a restaurant search app, which may have provided one or two suggestions ten years ago, and now returns all available options, highlighting the steady advance of biometric algorithms. In Spring of 2019, 30,000 unknown prints were submitted for rechecking with the Idemia algorithm by state police, and 11,000 new potential matches were returned. Further, 1,500 of those were for high-priority and violent crimes.
While the previous software returned matches for 20 to 30 percent of submitted prints, the current software generates a hit with nearly every print, according to the report.
When a match is produced, at least two MSP computer analysts review and verify the results.
The Intercept, meanwhile, reports that a trio of public defenders from Cook County, Illinois, with only the knowledge they had gained as criminal defense attorneys, aced a test for fingerprint examiners.
Though they had correctly identified all of the fingerprints on the test except for one, they were disconcerted that the test could be passed without any additional training, and wondered what that meant about the competence of fingerprint examiners such as the six working for the Chicago Police Department. Analysts had appeared to them to know very little about their discipline in cross-examinations, and the lab they work in does not follow any written policies or quality assurance practices designed for forensic processes, according to the Intercept.
Federal records show that 98 percent of practitioners in accredited public crime labs or certified by proficiency testing, but what level of proficiency do the tests really indicate? A 2009 report from the National Academy of Sciences suggested that pattern-matching disciplines, as opposed to DNA analysis, lack meaningful scientific underpinning.
The article suggests that the uniqueness of fingerprints has never been proven, and registers uncertainty over the impact of “noisy” data. Close non-matches have led to innocent people becoming suspects, as in the suspicion of Oregon Lawyer Brandon Mayfield in the 2004 Madrid train bombing.
Following the National Academy of Sciences report, some practitioners have made little change, while others have worked towards systematizing error rates for the field, investigated the causes of errors, and researched objective techniques.
A 2016 report by the President’s Council of Advisors on Science and Technology provided details on a pair of studies, one of which showed error rates up to 1 in 24, and recommended informing juries of the research.
The Cook County attorneys had heard police testify to scientifically unsupportable certainties, such as that a fingerprint match eliminated all other possibilities in the world, and found such testimony continued despite the evidence.
The passage rates for fingerprint proficiency tests conducted by Collaborative Testing Systems (CTS), a leading forensic testing material provider, are in the mid to high 90s each year. The test the attorneys took consisted of 12 complete, clean prints, without the noise or close non-matches that must be dealt with commonly in forensic analysis.
A round of testing in 1995 did include a couple such challenges, and less than half of those who took the test got all seven correct. The test was changed. The article notes that some practitioners have been complaining about the erase of the CTS test for years. Lab directors, however, do not want to pay $340 each for many tests if many staff will not pass.
CTS President Chris Czyryca notes that it is not really possible to replicate the conditions of forensic case work in a test which must be uniform for hundreds of people, and also says that getting only one question wrong does not indicate high proficiency the same way it might for other tests or subjects.
Current research by RTI International Research Scientist and Latent Print Examiner Heidi Eldridge to measure the quality of prints objectively may result in more nuanced, and therefore useful, tests.
The National Academy of Sciences report recommended all crime labs be independent, but the Houston Forensic Science Center is one of few in the country not associated with a police or prosecutor agency.
For its part, the Chicago Police Department says the quality of its fingerprint examiners will be assured by having them take an annual proficiency exam.
Article Topics
accuracy | biometric database | biometric software | biometric testing | biometrics | fingerprint recognition | forensics | IDEMIA
Comments