Mike May 2016-03-26 02:08:03
THE FUTURE OF IDENTIFYING PEOPLE WILL REQUIRE MORE THAN ONE METHOD. Standing in the immigration line at the Indira Gandhi International Airport in Delhi, India, I watch person after person be fingerprinted. First, you put your left four fingers on a digital pad, then your right four, and finally both thumbs at once. If all goes smoothly, the Indian government collects all ten fingerprints for everyone entering the country. It’s not as easy as it sounds, even from the start. More than one person is asked to try again and again. So obtaining a print can be as difficult as analyzing one. Although some fingerprint analysis is new, the concept— using fingerprints for identification—started centuries ago. Thousands of years ago in Babylon, a fingerprint served as a signature of sorts on business papers. Finally, in 1880, British surgeon Henry Faulds described using fingerprints to identify people; he gets credit for the first use of this technology of lifting a print from an alcohol bottle. By 1946, the U.S. Federal Bureau of Investigation (FBI) maintained a collection of 100 million prints, all kept on cards and maintained manually. In 1999, the FBI launched its computer-based Integrated Automated Fingerprint Identification System (IAFIS). This unique and largely unchanging feature, a fingerprint, now gets used more than ever—from criminology to home computing. For the field of criminology, Anil K. Jain—a computer scientist at Michigan State University in East Lansing and the creator of many tools for analyzing fingerprints, including AltFingID, which can detect altered prints— says, “Fingerprints are primarily for identifying repeat criminals.” At a crime scene, this can be challenging because, as Jain notes, “They can be partial prints—maybe just one or two fingers and smudgy.” Fingerprints can be visible or invisible. The former could be a fingerprint left in blood, for example, and this can be captured with a high-resolution picture. Invisible or latent prints need to be found and made visible. For latent prints, Eberhard Schultheiß, manager of R&D at Germany-based EVISCAN, says, “A large collection of methods exists to accomplish this, and such methods have in common that they use one- or two-step processes based on the use of chemicals.” He adds, “These chemicals are applied to either mask chemical substances left as latent traces, to stain such traces, or both.” Then, the stains used can be imaged under visible or fluorescent light. A range of technologies is applied to obtaining and analyzing fingerprints. The capture, processing, and analysis all impact the final results. Likewise, the technologies involved range from imaging to computation. Moreover, the world’s increasing population, easier access to travel and expanded use of fingerprints to identify users of personal electronic devices increase the need to identify fingerprints quickly and accurately for security reasons that range from public to personal safety. IMPERFECT IDENTIFICATION Although fingerprints do not change naturally with time, they can get worn, and that can make identification tricky. “Some people have poor-quality fingerprints for occupational reasons, like a mason doing brickwork,” Jain says. “It can be more difficult to identify a person who works with lots of chemicals too.” So there will always be some individuals who cannot be recognized with fingerprints. Beyond the fingerprints themselves, the capture technology creates some challenges. “The biggest challenge is that not all fingerprints that are left on different materials or surfaces can be developed easily,” says Kok-Jhoon (KJ) Wong, national sales manager at Australia-based Pathtech. “Prints left on samples that are later wetted, [on] polymer currency notes, or [on] porous surfaces can be difficult to develop with conventional methods.” The EVISCAN technology combines heat imaging and infrared spectral properties of the material. “When a heat imaging device is aimed at an exhibit surface where trace material is present, it will see differences in reflected infrared intensity between trace and exhibit—a contrast [that] is not seen in visible light,” Schultheiß explains. The EVISCAN excludes parts of the spectrum coming from the surface to detect latent traces without any need for contrasting chemicals or dyes. Once you have a print, it can be imaged in many ways. For example, Pathtech’s Foster + Freeman Crime-lite Imager has a full set of different wavelengths to capture fingerprints in different lighting conditions. This instrument is very easy to use and reduces the need for photographic expertise to capture a very good image. Where the print gets left, also matters. “Getting prints off bullets is very difficult, because when you fire a gun, any print that was left on the cartridge degrades by the heat generated,” Wong says. The CERA-LT, though, was designed specifically to recover fingerprints from bullet casings by using patented lighting and stitching software. “Normally bullets are loaded into the gun for a long time before its being fired. This gives sufficient time for the fingerprint residue—salt—to corrode the casing, leaving a mark. Hence it is possible to recover the print even after it is fired.” Other technologies also offer more options in the development of fingerprints. For example, vacuum metal deposition (VMD) uses metal ions, which can develop prints even on wetted samples. “In a wetted sample, only the oily residues of the fingerprint are left,” Wong explains. “VMD takes advantage of this oil residue and develops a negative image of the fingerprint.” MILLIONS AND MORE In analyzing fingerprints, quantity also matters. For example, the state of Michigan maintains a database of approximately three million, ten-finger sets of prints. So finding a criminal can mean trying to compare a partial print to 30 million possibilities. Some databases contain huge numbers of fingerprints. For example, the Unique Identification Authority of India (Aadhaar) will include fingerprints—all ten—and two iris images for each of the country’s 1.2 billion citizens. Today’s computer power, though, can run comparisons fast. If it’s just one print being compared to a thousand or so, says Jain, a personal computer or even a smartphone can do it in about a second. Even when comparing one print to a million prints, parallel processors make quick work of the analysis. “At border crossings or immigration centers at airports, where they collect ten prints from each visitor, it only takes a second or so to compare that to a watch list, because they are using efficient algorithms and distributed computing,” Jain explains. SPOILING THE SPOOFS With more and more countries adopting fingerprints for national identification programs and fingerprints being used to unlock mobile phones or to secure information, it raises the potential for a criminal finding a way around that. One of those methods is called spoofing, which is essentially mimicking a fingerprint. For example, one company made a fake fingerprint out of wood glue and easily tricked the fingerprint reader on a cell phone. In fact, a quick search online reveals more than one DIY spoofing video. Although this makes me feel less than secure as I open my new iPad with a touch of my thumbprint, experts are on the case. When asked about some of today’s biggest challenges with fingerprints, Stephanie Schuckers, director of the Center for Identification Technology Research (CITeR) at Clarkson University in Potsdam, New York, says, “People are committing fraud with fingerprints, faking a biometric device by trying to become someone else or to hide their own identity.” She adds, “If you are on a watch list, you have some incentive to hide your fingerprint.” Schuckers points out that commercial defenses against spoofing are emerging, but as she says, “the problem is not fully solved.” Schuckers works on ways to detect spoofed prints. “We do that through algorithms,” she says. “You look at small details in the image.” For example, an algorithm can look at grey-level information to try to distinguish a real from a fake fingerprint. Spoofers, though, create an ongoing battle: The software learns to identify a particular kind of spoof, the spoofers try to get around that, which drives the need for new or improved algorithms. Consequently, the details of many anti-spoofing algorithms remain proprietary. The future of identifying people will probably include some combination of methods. For instance, Schuckers is part of the FIDO Alliance, which stands for “fast identity online” and aims to reduce our reliance on passwords. The plan is to combine cryptography, such as a public-private key code, with a biometric identifier, such as a fingerprint. “The biometric communication is local, at the device,” she says, “and the cryptography is between the device and another party, such as a payer.” She adds, “This might not make us completely free of passwords, but it might limit their use to things like identity recovery.” CUTTING THE COST Seeing fingerprint readers showing up in mobile devices suggests the decreasing price of the technology. “The cost of the print reader in a mobile phone is only a couple of dollars,” Jain says. Inexpensive readers can mean lower quality. “Poor quality translates into poor performance in an algorithm,” says Schuckers. “In mobile devices, the price of the scanner is under pressure to go down, and that impacts the quality and the amount of information that you are able to capture.” For example, smaller scanners capture a portion of the print, and even that is at a lower resolution. “The sensor is being taken over by the consumer-device market,” Schuckers says. “How can we capture the fingerprint in an inexpensive way to put it on a mobile device?” A lower price, though, is not always good. “If you spend more money to get a better quality scanner,” says Schuckers, “the result is more robust.” Even though today’s print readers remain imperfect, that’s no reason to abandon them. As Schuckers says, “Even though whatever mechanism you put in place might not be perfect, like virus-protection software, you wouldn’t want to go without it.” Mike May is a freelance writer and editor living in Ohio. You may reach him at firstname.lastname@example.org.
Published by LabManager. View All Articles.