Author Archive Boyer

ByBoyer

Technology; Faster, More Accurate Fingerprint Matching

Automated fingerprint technology has been in development for years. But an industry began to emerge only when the Federal Bureau of Investigation committed itself in 1990 to convert its 40 million fingerprint cards and 24 million crime history records into paperless computer files by the end of the decade. To do so means police must be able to submit inquiries electronically.

“We think there will be a revolution in fingerprinting,” said David F. Nemecek, a top deputy for the F.B.I.’s information services division. In the forefront of the revolution are three companies that make live-scan machines. Last year, the F.B.I. tested all three technologies and ultimately approved them. Since then, about 100 local police agencies have installed live-scan machines; the biggest systems are Chicago’s, with 34, and Philadelphia’s 11.

There are two principal pieces of technology for automating fingerprints: an electronic reader or live-scan machine, costing up to $60,000 each, and a “matching” computer, which compares new fingerprints with prints on file. Matching computers, which can cost up to $30 million each, are already installed in 36 states and process fingerprint cards that are scanned into the system. When live-scan machines are in use, this process will be greatly speeded up.

But being able to scan fingerprints into a computer is only part of the process of matching prints. Human technicians classify each print by subgroups based on the century-old Henry system of whorls, arches and loops. Then the computer digitizes an image by creating a map of the unique ridge patterns, which are called minutiae points.

The computer translates 90 or more minutiae points per finger into a binary code. Then, the computer attempts to make a match of minutiae points by following a set of instructions called a searching algorithm.

The computer is not actually comparing fingerprint images; it is trying to find a group of prints with similar binary minutiae codes. A technician makes the final match from a candidate list. It is possible to make a match with as few as eight minutiae points.

In live-scan electronic fingerprinting, a suspect’s finger is placed on a glass plate and its image captured with a video scanner at a precise oblique angle. Light that is completely absorbed is seen as black, and light that is completely refracted is seen as white. To the video scanner, finger ridges turn black and valleys white.

As the finger is rolled across the plate, the computer builds an image over time, and the image is digitized by its light value into binary code and entered into the memory.

The emerging business of producing live-scan machines could explode to $750 million in sales in five years from present sales of under $20 million, said Carol M. Clark, research director for John G. Kinnard & Company in Minneapolis. Besides the nation’s 16,000 law enforcement agencies, potential buyers include companies that fingerprint their employees as well as the military and the Internal Revenue Service.

Digital Biometrics Inc., based in Minnetonka, Minn., and Identix Inc. of Sunnyvale, Calif., use similar video optics devices to capture the fingerprint image. Fingermatrix Inc. of White Plains, N.Y., uses a different optical device and a U-shaped platen. Its sensor moves around the platen.

There are other slight differences in the competitors’ machines. Digital Biometrics’s system has live video feedback so the booking officer can see the finger on a monitor as it is rolled. Identix and Fingermatrix machines replay the captured image a second later, although Identix ultimately expects to provide real-time displays.

As a booking officer takes prints, they are captured into a single computer file, which also includes biographical information, and are then stored on the computer’s hard disk. The file can then be reproduced on an F.B.I. “tenprint” card or sent electronically by a high-speed modem to a printer at a state or Federal crime computer.

“Ultimately, police want to never print it out but keep it in digital form,” said Daniel F. Maase, Identix’s engineering vice president. So far that is not possible because no one has created a computer that can automatically classify prints into Henry subgroups. The Orincon Corporation in San Diego is working on that puzzle with neural networks.

Depending on the live-scan equipment used, a single fingerprint file can amount to 1.5 million to 2.5 million bits of information. Live-scan machines can send files six times faster than the standard electronic transmission speed, but even so, transmission can take up to 14 minutes per card. Prospective live-scan vendors to Los Angeles County will have to evaluate and propose the best transmission scheme for their systems.

Live-scan prints should markedly improve law enforcement’s weakest link. The F.B.I. rejects 18 percent of the ink-and-roll cards sent by local agencies. More than half of the eight million fingerprints in California’s matching computer are of poor quality. It is easy to see why. Suspects get impatient as jailers collect three sets of originals for local, state and Federal files. Typically, the F.B.I. gets the worst of the three.

The next step for live-scan manufacturers is making a single-finger mobile scanner for use in patrol cars. Officers could then identify a fugitive by transmitting a suspect’s print to the F.B.I.’s computers from the street.

Prototypes have already been shown. “It’s not a dream,” Mr. Nemecek said. “We found the technology was available, and we obtained the kind of accuracy we wanted.”

Bids are under review, he said. An award is to be made in December, and mobile scanners, along with small printers, are to be installed in some F.B.I. cars next year.

ByBoyer

Your fingerprints aren’t as unique as a snowflake. Yet.

Governments and law enforcement bodies are searching for better fingerprint capture technology, and the results of a recent contest demonstrate they may be getting closer to a solution.

A report by the National Institute for Standards and Technology details results from a September 2017 challenge held by the Intelligence Advanced Research Projects Activity to find more accurate (and less uncomfortable) method to capture fingerprints.

The project builds on previous work by IARPA over the past year to improve fingerprint analysis through better technology. In 2017, the agency stood up two programs: one called Odin, for detecting fake or altered fingerprints, and another, Nail to Nail [N2N], a form of contact less fingerprint analysis that has become the standard the agency is hoping to best through its prize challenge.

While the N2N program did develop a way to more accurately capture fingerprints compared to lower-tech plain impressions, it’s not exactly user friendly. In order to get an accurate capture, the operator must physically hold a subject’s finger while repeatedly rolling it on a live scan platen, a process that NIST describes as “time consuming” and “uncomfortable.”

“Compared to [Nail to Nail] impressions, plain impressions capture a limited region of the finger, resulting in less information being made available for searching with automated fingerprint identification algorithms,” the report states. “This lack of searchable information is especially detrimental in forensic applications because latent fingerprints found at crime scenes are sometimes formed from areas of the finger not imaged by a plain impression.”

That led IARPA officials to set up the challenge, canvassing eight organizations from industry and academia to develop better sensor technologies that are more seamless than N2N and perform better than plain impressions. The challenge had four separate prizes: one for logging the fastest N2N acquisition time, two for the device with the most accurate image capture and a grand prize for the best overall identification system.

Two companies – Advanced Optical Systems and Green Bit – took home the first three prizes, each worth $25,000. However, none of the solutions wound up coming close enough to match the accuracy of N2N images and as a result, no one took home the $100,000 grand prize.

The urgency for better fingerprint analysis is real. While popular culture has pushed the idea that everybody’s individual fingerprint is as unique as a snowflake, forensic scientists have increasingly raised questions about the accuracy of current technologies.

Last year the American Association for the Advancement of Science released a report concluding that any courtroom testimony or reports stating or implying that fingerprints collected from a crime scene belong to a single person are “indefensible and lack scientific foundation.” This was particularly true for latent fingerprint analysis, one of the metrics IARPA tracked through its N2N challenge.