Each person’s fingerprints are unique, even those of identical twins, making them a useful tool for identification. And fingerprints aren’t the only unique markers—the field of biometrics has expanded in recent years to include identification through voice, the irises and retinas of the eyes, gait, and DNA.
Now, an algorithm developed by researchers at Jadavpur University in Kolkata, India, has added facial blood vessels to the list.
The study, published by Ph.D. candidate Ayan Seal in the upcoming issue of the International Journal of Computational Intelligence Studies,
used an infrared thermal camera to read the heat signature generated by
capillaries lying just under the skin of the face. Capillaries are
extremely small blood vessels, some so narrow that red blood cells have
to travel through them single-file.
Although major structures like arteries tend to appear in more or less the same place in each person, capillaries branch off of larger vessels as needed. Thus, the network of capillaries in each person’s face is completely unique.
“The thermal imprints of the blood vessels may be treated as the ridges in fingerprints, and fingerprint recognition techniques may be applied on thermal imprints of human faces for their recognition,” Seal told Healthline. “A computer can rapidly compare this function with that of anyone else in the world whose face image has been scanned.”
Seal created a baseline image for each person he studied by taking 39 different thermal images from a fixed distance, with different facial expressions, in different poses, and with the face partially covered by the hands.
More Than Skin Deep
In addition to identification, Seal’s technology may also have medical applications.
“There are some utilities for reconstructive surgery,” Dr. Hooman Khorasani, Chief of Dermatological and Cosmetic Surgery at Mount Sinai Medical Center, told Healthline. “For instance, part of the nose may need to be constructed if cancer took it. You lose skin, muscle, cartilage. You find a way to take skin from somewhere else in the body, say the forehead. Now, the thing is, that the flap needs to bring in its own blood supply, or it may not survive.”
In order to correctly graft tissue, Khorasani must map where its blood supply is coming from and where he will hook up the arteries and veins at the transplant destination. If he's transplanting muscle tissue and also needs to make sure he can connect up the nerves, precision becomes crucial.
Currently, Khorasani and other surgeons image soft tissue using high-definition ultrasound. The problem with ultrasound is that the machines are large, not very portable, and relatively uncommon. If all someone needed was an infrared camera, imaging could become a simple matter of taking a snapshot while a patient is in the office.
Since infrared light
operates on much smaller wavelengths than ultrasound, in theory, it
should be capable of much finer image resolution. However,
high-definition ultrasound is already a very advanced technology, and
Khorasani wonders if infrared could catch up.
“With high-definition ultrasound, nowadays you can even pick up skin cancer. These were things people never thought were possible before," Khorasani said. "Being able to pick up soft tissue structures that are so small, like a hair follicle. Not just the hair itself, but where the muscle attaches, where the nerves attach.”
Ultimately, Khorasani says, whether Seal’s technology will enter widespread medical use will depend on its resolution and price. As a surgeon at a research institution, he says, he gets to do clinical trials for new devices all the time, but most of them never leave the lab because they prove too expensive for his colleagues in private practice to afford.
Is Big Brother Watching You?
Seal’s algorithm currently doesn’t work on moving subjects and requires
a person’s cooperation to establish a baseline, huge advances in facial
recognition software suggest these stumbling blocks may be short-lived.
One advantage of thermal imaging is that, unlike visual imaging, it can work in complete darkness and if a person is wearing a mask. Thermal cameras are currently quite expensive, but that too will change over time.
Facial recognition, like eye scans, can be used to identify a person without his or her knowledge. These methods raise privacy concerns for people like Jennifer Lynch, staff attorney at the Electronic Frontier Foundation.
“If someone is walking around in public and a camera can identify them at a distance, that's problematic as a First Amendment issue—we should be able to engage in protest anonymously,” Lynch said in an interview with Healthline. “We've seen that tracking chills people's tendency to associate. It makes people not want to hang out with their friends if they're worried about participating in activities with groups that tend to be on the fringe.”
For such a system of remote identification to work, the government would have to routinely collect thermal images of its citizens, as the United States currently does with fingerprints and the UK does with DNA, in order to have a database of images to check against. Creating such a database presents its own problems.
“If it's a system run by the government, that means that the government has a ton of information about the populace of the country," Lynch said. "The government shouldn't have the ability to surveil people. If the system is used by, say, a company to verify entry into a building, it's a much smaller database—it's only a database of employees at the company, so it doesn't create the same kind of privacy risk as a much larger system.”
Now You See Me...
Although it's very difficult, it is possible to create a cast of false fingerprints or a contact lens that mimics the patterns of the iris. Facial capillaries, on the other hand, would be impossible to forge.
In Seal’s experiment, his algorithm proved to be 97 percent accurate. But is that accurate enough?
“One of the biggest issues with biometric identification systems is reliability. Too many false positives, too many false negatives,” security professional Kurt Narveson told Healthline. “Any time you use it for a security standpoint, there have to be no false positives. False negatives are annoying, but can be dealt with, but a false positive is a breach.”
However, he says that a multimodal approach—say,
requiring someone to pass a scan and then enter a PIN number—could
compensate for accuracy. Although thermal face scanning is unlikely to
be used as access control for classified government material, he
suggests it might have applications elsewhere, for example, at banks.
As Narveson says, “If you're going to rob a reserve bank of millions of dollars, you don't show up with a gun—they have 30 guys with guns. You show up in a business suit with access to their system. If that's a slide card, that's much easier to beat than an image of your face. You can pickpocket a card off someone, but you can't pickpocket their face.”
However, Jennifer Lynch has her own concerns about Seal’s security claims. “I haven't seen a biometric system yet that couldn't be spoofed in one way or another,” she said. "It always worries me with any system when a company says it's invulnerable to spoofing or hacking; that's a big red flag. It just says they haven't hired the right people to try to spoof their system yet.”
At the end of the day, Narveson agrees. “The best system? Personal recognition. The human brain picks up on far more information than we can measure or are consciously aware of," he said. "It's all about ‘this doesn't seem correct.’”