Dear Friends,
Be Well.
David
11 Body Parts Defense Researchers Will Use to Track
You
BY NOAH SHACHTMAN AND ROBERT BECKHUSEN01.25.136:30 AM
Cell phones that can identify you by how you walk.
Fingerprint scanners that work from 25 feet away. Radars that pick up your
heartbeat from behind concrete walls. Algorithms that can tell identical twins
apart. Eyebrows and earlobes that give you away. A new generation of
technologies is emerging that can identify you by your physiology. And unlike
the old crop of biometric systems, you don't need to be right up close to the scanner
in order to be identified. If they work as advertised, they may be able to
identify you without you ever knowing you've been spotted.
Biometrics had a boom after 9/11. Gobs of government
money poured into face and iris recognition systems; the Pentagon alone spent
nearly $3 billion in five years, and the Defense Department was only one of
many federal agencies funneling cash in the technologies. Civil libertarians
feared the worst as face-spotters were turned on crowds of citizens in the
hopes of catching a single crook.
But while the technologies proved helpful in verifying
identities at entry points from Iraq to international airports, the hype -- or
panic -- surrounding biometrics never quite panned out. Even after all that
investment, scanners still aren't particularly good at finding a particular
face in the crowd, for example; variable lighting conditions and angles (not to
mention hats) continue to confound the systems.
Eventually, the biometrics market -- and the
government enthusiasm for it -- cooled off. The technological development has
not. Corporate and academic labs are continuing to find new ways to ID people
with more accuracy, and from further away. Here are 11 projects.
Above:
The Ear
My, what noticeable ears you have. So noticeable in
fact that researchers are exploring ways to detect the ears' features like they
were fingerprints. In 2010, a group of British researchers used a process
called "image ray transform" to shoot light rays at human ears, and
then repeat an algorithm to draw an image of the tubular-shaped parts of the
organ. The curved edges around the rim of the ear is a characteristic -- and
most obvious -- example. Then, the researchers converted the images into a
series of numbers marking the image as your own. Finally, it's just a matter of
a machine scanning your ears again, and matching it up to what's already stored
in the system, which the researchers were able to do accurately 99.6 percent of
the time. In March of 2012, a pair of New Delhi scientists also tried scanning
ears using Gabor filters -- a kind of digital image processor similar to human
vision -- but were accurate to a mere 92 to 96.9 percent, according to a recent
survey (pdf) of ear biometric research.
It may even be possible to develop ear-scanning in a
way that makes it more reliable than fingerprints. The reason is because your
fingerprints can callous over when doing a lot of hard work. But ears, by and
large, don't change much over the course of a lifespan. There's a debate around
this, however, and fingerprinting has a much longer and established history
behind it. A big question is whether ear-scanning will work given different amounts
of light, or when covered (even partially) by hair or jewelry. But if
ear-scanners get to the point of being practical, then they could possibly work
alongside fingerprinting instead of replacing them. Maybe in the future we'll
see more extreme ear modification come along as a counter-measure.
Photo: Menage a Moi/Flickr
Odor
In the early and mid-2000s, the Pentagon's blue-sky
researchers at Darpa dabbled in something called the "Unique Signature
Detection Project," which sought to explore ways to detect people by their
scent, and maybe even spot and identify individuals based on their distinct
smells. Darpa's work ended in 2008. The following year, the Department of
Homeland Security fielded a solicitation for research in ways that human scent
can indicate whether someone "might be engaging in deception,"
specifically at airports and other ports of entry.
Odor detection is still just a research project at the
moment. The science is intricate, involving more than 300 chemical compounds
that produce human odor. Our personal stinks can change depending on everything
from what we eat to our environment. But it may be possible to distinguish our
"primary odor" -- separate from "secondary" odors based on
our diet and "tertiary" odors based on things like soaps and
shampoos. The primary odor is the one linked to our genetics, and there have
already been experiments with mice, which have been found to produce distinct
scents unique to individuals. In 2007, the government's counter-terror
Technical Support Working Group even started a program aimed at collecting and
storing human odors for the military's dog handlers. Dogs, of course, have been
used to track people by smell for decades, and are believed to distinguish
between humans based on our genetic markers.
Photo: Cabaret Voltaire/Flickr
Heartbeat
Your chest moves, just a little, every time your heart
beats or your lungs take in air. For years, researchers have been monkeying
with radars that are sensitive enough to to detect those minuscule chest movements
-- but powerful enough to do it from hundreds of yards away. Even reinforced
concrete walls and electromagnetic shielding won't stop these radars, or so
claim the researchers at the small, Arizona-based defense contractor VAWD
Engineering, who are working on such a system for Darpa's
"Biometrics-at-a-distance" program.
The key is the Doppler Effect -- the changes in
frequency when one object moves relative to another. We hear it all the time,
when a fire engine passes by, siren blaring. VAWD says their vehicle-mounted
Sense Through Obstruction Remote Monitoring System (STORMS) can pick up even
small fluctuations of chests.
STORM (pictured above) "can be used to detect,
classify and identify the specific cardiac and pulmonary modulations of a...
person of interest," a company document boasts. By itself, a heartbeat or
a breathing rate won't serve as a definitive biometric. But combine it with
soft biometrics (how someone subtly sways when he or she stands) and you've got
a unique signature for that person that can't be hidden or covered up.
VAWD says these signature will help improve disaster
relief and medical care by providing a "reliable, real time medical status
equal to or better than the current devices, while increasing the mobility and
comfort of the patient."
But the company also notes that its system performs "automated
human life-form target tracking" even when construction materials like
"Afghan mud-huts" are in the way. STORM "has already been
deployed by the United States Army on one of its most advanced ground
vehicles," the company adds.
Does any of that sound like hospital work to you?
Illustration: Yale University/Wikimedia
Photo: VAWD Engineering
Voice
Most people are likely to be familiar with voice
readers on gadgets like the iPhone. But what if there was software that could
quickly analyze the voice of thousands, and even use those voices to identify
specific people?
Russian biometrics firm Speech Technology Center --
known as SpeechPro in the U.S. -- has the technology. Called VoiceGrid, the
system is able to automatically recognize a person's voice as their own,
provided your voice is pre-recorded in a database and can be recalled by the
computer. The company has also developed a version for "large city,
county, state or national system deployments."
It's seen use in Mexico, according to Slate,
"where it is being used by law enforcement to collect, store, and search
hundreds of thousands of voice-prints." The National Security Agency has
taken interest in similar technology. So has the FBI. A 2012 presentation from
the National Institute of Standards and Technology -- with the assistance of
the FBI -- also speculated on potential uses including identifying and clearing
people 'involved in illegal activities," locating serial killers and
identifying arms traffickers (.pdf). Iarpa, the intelligence community's
research agency, has also been looking into ways to solve some of its problems:
audio interference mainly. In 2011, the agency concluded its Biometric
Exploitation Science and Technology Program (or BEST), which made "speaker
recognition advances which included improving robustness to noise,
reverberation, and vocal effort, and by automatically detecting these
conditions in audio channels," spokesperson Schira Madan told Danger Room
in an email. But we wonder if it'll detect autotune.
The Iris
Imagine a scanner than can look deep inside your eye
-- from 10 feet away. Actually, you don't have to think that hard. The
technology is already here. Scanners have been developed that can focus in and
scan irises from a distance of 10 feet, such the IOM PassPort, developed by
government contractor SRI International. The company promises the machine can
scan irises at a rate of 30 people per minute -- like in high-traffic areas
such as airports and train stations. SRI also claims it can see through contact
lenses and glasses.
But the longer-range scanners could also see other
uses, aside from airports. U.S. troops field existing, short-range and handheld
iris scanners to build databases of Afghan eyes as part of a plan to use
biometric data to tell civilians apart from insurgents. The Department of
Homeland Security has tested iris scanners at a Border Patrol station along the
Texas-Mexico border. The FBI has been working on an iris database for federal
prisoners, and Google uses them at company data centers. But these systems can
be fussy, and require that the targets don't move too much.
There might be another way. The Pentagon's scientists
at Darpa have funded a research project at Southern Methodist University to
develop cameras that can automatically zoom-in and scan irises, kinda like what
happened to Tom Cruise in Minority Report -- and without being blocked by pesky
obstructions like eyelashes and glare from light. But another problem is that
iris scanners are not the most secure means of identifying people. In July
2012, a group of researchers from the U.S. and Spain discovered a way to spoof
the scanners by duplicating iris images stored in databases and creating
synthetic copies. That means someone could conceivably steal your eyes, in a
way.
Illustration: Air Force
Periocular
Spotting someone by their irises is one of the
best-developed biometric techniques there is. But Savvides and his Carnegie
Mellon colleagues think there may be an equally-promising approach in the area
around the eye -- also known as the "periocular" region.
The "periocular region has the most dense and the
most complex biomedical features on human face, e.g. contour, eyelids, eyeball,
eyebrow, etc., which could all vary in shape, size and color," they wrote
in a 2011 paper. (.pdf) "Biologically and genetically speaking, a more
complex structure means more 'coding processing' going on with fetal
development, and therefore more proteins and genes involved in the
determination of appearance. That is why the periocular region should be the
most important facial area for distinguishing people."
And unlike other biometrics -- the face, for instance
-- the periocular region stays remarkably stable as a person ages. "The
shape and location of eyes remain largely unchanged while the mouth, nose,
chin, cheek, etc., are more susceptible to changes given a loosened skin,"
the researchers note. In other words, this is a marker for life.
Nearby, Savvides and his colleagues think they've
found a second biometric: the shape of the eyebrow. Face-scanners are sometimes
thrown off when people smile or frown. But the eyebrow shape is
"particularly resilient to certain (but not all) expression
variations," the researchers note in a separate, yet-to-be-published
paper. And the eyebrow can still be seen, even when the subject has most of his
or her face covered.
What's not fully clear is how the eyebrow biometric
responds to threading, shaving or waxing. Saavides, who responded to tons of
questions about his research, says there's no fullproof means to avoid this
kind of spoofing. But Saavides is also working on sensors that can analyze
multiple facial cues and features, while incorporating algorithms that detect
the possibility of a person changing one or two of them. A pair of plucked
eyebrows might be a weak match compared to the bushy ones the computer has on
file -- but the computer could also be smart enough to recognize they've been
plucked.
Photo: Carnegie Mellon University
Long-Range Fingerprint Scanners
Most fingerprint scanners today require physical
contact, but constantly being soaked with finger-oil and dirt can also muck-up
the machines. For that reason, among others, one developer is working on a
scanner that may one day read your fingerprints at a distance of 20 feet.
But first, scanners with a 20-foot distance haven't
hit the market quite yet. One machine called the AIRprint, made by Alabama firm
Advanced Optical Systems, has a range of nine feet, and uses two 1.3 megapixel
cameras that receives light in different wavelengths: one horizontally
polarized, and the other vertically polarized. To sort out the different
wavelengths, a device beams light at your fingerprints, which bounce back into
the lenses, which then combines the separate wavelengths into a clear picture.
A spin-off company called IDair also has a commercial scanner that reaches up
to six feet and is marketed toward "security personnel." IDair's
20-foot-range machine is currently in development, and is described as
functioning similar to satellite imagery.
The military is reportedly an interested customer. The
MIT Technology Review surmised that Marines may use them for scanning
fingerprints from inside the relative safety of an armored vehicle or behind a
blast wall. It beats exposing yourself to the possibility of a suicide bomb
attack. For the civilian market, that seems better than pressing your
fingertips against a greasy scanner, if you're comfortable with the idea of
having your prints scanned from far away.
Photo: LetTheCardsFall/Flickr
Gait
Even before 9/11, researchers were floating that
notion that you could pick out someone by how he or she walks. And after the
Towers fell, Darpa made gait recognition one of the cornerstones of its
infamous Total Information Awareness counterterror program.
The problem is that gait can be kind of hard to spot.
A briefcase or a bum leg prevents the recognition system from getting a clear
view. So filming someone walk didn't make for a particularly reliable biometric
system. Plus, the same person may have multiple gaits -- one for walking, and
another for running, say.
But the spread of smartphones has opened up a new way
of identifying someone's stride. Androids and iPhones all have accelerometers
-- sensors that measure how far, how fast, and with how much force an object
moves.
"By using the accelerometer sensor in the cell
phone, we are able to capture a person's walking pattern. As it turns out,
these patterns are very good biometric traits for people identification. Because
it does not require any special devices, the gait biometrics of a subject can
even be captured without him or her knowing," write Carnegie Mellon
University professor Marios Savvides and his colleagues in a recent paper.
(.pdf)
In a small, preliminary study, Savvides and his fellow
researchers at the CyLab Biometrics Center claim they were able to get a 99.4%
verification rate with the system when the subjects were walking. 61% of the
time, they were even able to match someone's fast-paced gait to their slower
one. In other words, you can run.... but with a phone in your pocket, it's
going to be harder to hide.
Photo: sfllaw/Flickr
Sweat
The Army wants to see some sweat. No, not workout
sweat, but sweat that can betray hostile intentions. In 2010, the Army awarded
a nearly $70,000 contract to California security firm Irvine Sensors
Corporation to develop software that can use sensors to recognize at
"abnormal perspiration and changes in body temperature." The idea was
to determine "harmful intent in such military applications as border patrol,
stand-off interrogation, surveillance and commercial applications"
including surveillance at businesses and "shopping areas." It's a bit
out there, and still very much in the research stage, but makes a certain kind
of sense. Elevated stress levels could give a suspect away when scanned by
hyperspectral sensors that read changes in body temperature.
Though a reliable system will have to work in
combination with other biometric signals: threatening body movements, facial
expressions, iris scans -- all of these will also have to be factored into
determining whether someone is up to no good. The Army contract, dubbed Image
Analysis for Personal Intent, also sought to develop sensors that read these
signs from a distance of nearly 150 feet. Perhaps a bit optimistic. But in
2002, a group of scientists in Minnesota managed to determine if military
recruits were engaging in deception by scanning for changes in temperature
around their eyes. So if you're at all freaked out about the idea of
sweat-scanners, now might be time for a cold shower.
Photo: Army
Advanced Face Recognition
Most machines that scan and recognize your face
require taking a good, clean look. But now researchers are working on replacing
them with scanners that only need a few fragmentary snapshots at much longer
ranges than ever before.
One machine that can do it is being developed by
defense contractor Progeny Systems Corporation, called the "Long Range,
Non-cooperative, Biometric Tagging, Tracking and Location" system. Once a
person of interest is spotted, the system captures a 2D image of the person's
face before converting it into 3D. Then, once the image has been converted and
filed in a database, it can be quickly recalled when the system spots the
person for a second time. That's because 3D reduces the number of pixels needed
to analyze the image, speeding up the process and allowing the system to
identify a person with a mere glance. The company also claims the machine works
at more than 750 feet.
But a face alone may not be enough to recognize a
person with a machine. Everything from lighting conditions to distance can make
it harder to get a clear picture, especially if the person being scanned is on
the move, in a crowd, or ducking in and out of buildings. Using 3D models makes
it easier, but the technology will likely have to be combined with "soft
biometrics" like an individual's gender, height, weight, skin color and
even tattoos.
Slightly creepy, no? Well, it gets creepier, like the
group of Swiss scientists working on scanning facial features to detect your
emotions. Developers at Carnegie Mellon University have also developed a mobile
app called PittPatt --which has since been acquired by Google -- that can scan
your face and match it up with images you've shared over the internet, all in
less than a minute.
Photo: Carnegie Mellon University
Rapid DNA Testing
It used to be that DNA testing took months to perform,
from the time when a DNA sample was picked up on a swab, to analyzing it and
creating a DNA profile. Now there are machines that can do it in less than 90
minutes, and the Pentagon wants them.
This month, researchers at the University of North
Texas are beginning to test a $250,000 machine for the Defense and Justice
Departments, and the Department of Homeland Security, so that "casualties
and enemies killed in action can be quickly identified in the field,"
according to the Biometrics Research Group. But according to the October issue
of Special Operations Technology magazine, rapid DNA testing systems
co-developed by defense giant Northrop Grumman had already been delivered to "unspecified
government customers" beginning back in August. One of those customers is
believed to be the FBI. California company IntegenX also has a portable
rapid-DNA machine that can analyze molecules taken off everything from clothing
to cigarette butts. There's a simple reason why police are so interested. For a
burglar who's breaking into houses and leaving a DNA trail, the machines could
clue-in faster than the burglar is able to continue the spree.
Photo: US Navy
No comments:
Post a Comment