Christopher Mims reports in the Wall Street Journal:
Teaching phones what our faces look like will be just like teaching them our fingerprints. An image of your face is captured, relevant features are extracted and the phone stores them for comparison with your face when you unlock the phone. As with fingerprint recognition, the facial images are securely stored only on the device itself, not in the cloud. “Face recognition is has the ability to be much more ubiquitous than fingerprint scanning because camera sensors are that much easier to deploy.”
It was a memorable moment in Pixar’s 2004 classic “The Incredibles,” one that seemed wildly futuristic at the time: Mr. Incredible picks up a wafer-thin tablet computer, and it scans his face to verify his identity before divulging his secret mission.
Thirteen years later, many slim phones and tablets unlock with the press of a thumb—and just this sort of mobile facial scanning is on the way.
Forget fiddling with passwords or even fingerprints; forget multiple layers of sign-in; forget credit cards and, eventually, even physical keys to our homes and cars. A handful of laptops and mobile devices can now read facial features, and the technique is about to get a boost from specialized hardware small enough to fit into our phones.
Using our faces to unlock things could soon become routine, rather than the purview of spies and superheroes.
Qualcomm Inc., QCOM -0.27% an industry leader in mobile device chips, recently introduced its Spectra imaging system, which can extract depth information from objects including faces. The company plans to include the technology in a forthcoming generation of its flagship Snapdragon mobile processors. Meanwhile, when firmware for Apple Inc.’s forthcoming HomePod speaker leaked online, developers spotted clues suggesting that an upcoming iPhone might have similar depth perception and facial recognition.
This technology is different from, but related to, the facial recognition increasingly built into security cameras around the world, which cross-references pictures of your face against databases of millions. That tech is growing in capability and in use—especially in China, where its applications range from surveillance to payments.
Fortunately, your phone’s camera has a few advantages over surveillance equipment. It doesn’t need to spot you in a crowd. It just needs to distinguish one face—yours—and it can do that very well, since you’re not some shadowy figure captured in bad light. From a foot or two away, your phone can quickly capture a detailed image.
There’s another advantage. Depth-sensing technology, generally called “structured light,” sprays thousands of tiny infrared dots across a person’s face or any other target.
By reading distortions in this field of dots, the camera gathers superaccurate depth information. Since the phone’s camera can see infrared but humans can’t, such a system could allow the phone to unlock in complete darkness.
While Apple hasn’t announced any use of this technology—let alone confirmed whether it will exist inside the widely expected 10th-anniversary iPhone—the company is no stranger to infrared depth-mapping and facial recognition. It has previously been granted patents describing nearly identical processes. Apple declined to comment on any technology it might be working on.
If this technology sounds familiar, it’s a sort of shrunken-down version of the Xbox 360’s Kinect motion sensor. Perhaps not coincidentally, Apple acquired an early Kinect developer, the Israeli startup PrimeSense, in 2013.
Meanwhile, Qualcomm says it plans to make its Spectra processor available for future Android phones. Previous Samsung image processors that did face recognition could be fooled by holding up a photo of someone’s face to a phone’s camera. Qualcomm insists that depth perception gives the added bonus of “live-ness detection.” As a result, a 3-D printed mask wouldn’t be able to fool the system, though the company admits identical twins might.
Teaching our phones what our faces look like will be just like teaching them our fingerprints, says Sy Choudhury, a senior director at Qualcomm responsible for security and machine-intelligence products. An image of your face is captured, relevant features are extracted and the phone stores them for comparison with your face when you unlock the phone.
As with fingerprint recognition, the facial images are securely stored only on the device itself, not in the cloud. History—from Apple’s battles with domestic law enforcement over unlocking iPhones to Amazon’s insistence that the Alexa doesn’t upload anything until it hears its wake word—suggests companies will use this privacy as a selling point.
Already, laptops use Microsoft’s Windows Hello face recognition for easier unlocking; some devices are equipped with Intel ’s RealSense 3-D depth camera, which preceded Qualcomm’s Spectra.
As technology like this gets into the mobile supply chain, it eventually trickles down to less-expensive, lower-end devices along with dedicated, highly efficient chipsets, says Joey Pritikin, founder and co-chief executive of biometrics company Tascent.
Facial recognition is likely one day to appear in camera-enabled smart doorbells and locks, as well as in smart speakers like Amazon’s camera-equipped Echo Show, where personalization would be a benefit: If it knew it was you, it might offer the latest “Game of Thrones” episode; if, instead, it spotted your child, there’s “Sesame Street.” Qualcomm already has customers in its Spectra pipeline working on nonphone products, including internet-connected cameras, says Mr. Choudhury.
Like all new technologies, facial recognition is likely to come with trade-offs. In many contexts, a fingerprint scan may be less obtrusive. The facial algorithms can be stymied by the presence or absence of glasses, especially sunglasses. The annoyance of having to take off your winter gloves to unlock your phone might be replaced by the annoyance of having to re-enroll your face if your facial hair changed.
All biometrics have their trade-offs—your irises may be even more unique than your face, but to scan them you have to bring the phone close to your eye. And since an iris lacks contour, it can still be spoofed by a still image. If you fail the facial scan, or if you’re wearing something that covers your face, you’ll have to punch in a code instead.
“The interesting thing about face recognition is that it has the ability to be much more ubiquitous than fingerprint scanning because camera sensors are that much easier to deploy,” Mr. Pritikin says. “I think it’s just a matter of time before our daily routine will reflect a number of seamless biometric authentications.”
Since Apple’s existing fingerprint sensor can already verify payments in stores via Apple Pay, it makes sense that Apple’s phones could also enable payments through face recognition in the future.
Perhaps the most remarkable thing about all this facial recognition is how mundane it has the potential to become, and fast. As a security measure that requires us only to look at our device, it’s easily taken for granted.
That is a good thing. Facial recognition has the potential to make higher levels of security—so-called multifactor authentication—so convenient that we no longer engage in all the bad habits that put our finances and online lives at risk. That isn’t to say these systems won’t be compromised somehow, but at least for the short term, smiling at your phone every morning will be far safer than punching in the same password over and over.
0 comments:
Post a Comment