facial recognition

Facial recognition technology scans faces, either in person or on a photograph, and measures distinguishing facial features such as eye position, eyebrow shape, and nostril angle. This creates a distinctive digital faceprint which the system runs through a database to check for a match.

Historical perspective: According to The Week, law enforcement agencies have had faces on file for decades; their databases provide them with the identified person’s name, address, age, and criminal history. By 2017 facial recognition was increasingly used by commercial firms too: Facebook’s system for tagging a photo, and Apple’s iPhone X can be unlocked when its owner simply looks at it.

Buy by 2018, more than two dozen AI experts signed a letter urging Amazon to stop selling its facial recognition technology to law enforcement agencies, said The New York Times. The experts, including a winner of the Turing Award—the Nobel Prize of computing—argued that Amazon’s "Rekognition" program is biased against women and minorities. Researchers at MIT found Rekognition had more trouble identifying the gender of female and darker-skinned faces in photos than similar services from IBM and Microsoft. It mistook women for men in almost one in five cases, and did even worse with darker-skinned women, getting the gender wrong 31 percent of the time.

As the technology becomes more widespread, there are growing fears that it will erode privacy and be misused by bad actors. Here are several FAQs to consider if a world of ubiquitous automated identification is one we really want to build.

Who uses facial recognition technology?
Facial recognition is most common in China, where people can use it to pay for a coffee, visit tourist attractions, and even withdraw cash from ATMs. Several Chinese cities use face-scanning cameras to shame jaywalkers, by flashing their names and photographs on public display boards. But the West isn’t lagging too far behind. In Europe, high-end hotels and retailers use facial recognition cameras to identify VIPs and celebrities as they enter, in order to give them preferential treatment. Several U.S. airlines are looking to replace boarding passes with face scanners. Department stores are using facial recognition to monitor how customers react to certain product displays. And these developments are only the tip of the iceberg.

What else is coming?
Doctors have already started using facial recognition to help them diagnose rare genetic diseases that produce distinctive facial characteristics; as the technology improves, they should be able to do the same for more common conditions, such as autism. Shops will soon be able to identify individual customers as soon as they walk in the store, and try to sell them specific items based on their interests and previous transactions. Dubai International Airport is scrapping one terminal’s security clearance counter altogether, and replacing it with a short tunnel fitted with 80 face-scanning cameras hidden behind video screens.

How are law enforcement agencies ramping up their facial recognition capacity?
Through various state and federal databases, the FBI now has access to photographs of half the U.S. adult population, according to a major 2017 report by the Georgetown Law Center on Privacy and Technology. Eighty percent of these people don’t have a criminal record; their faces are on file solely because they have some form of state ID, such as a resident’s card or a driver’s license. Several police departments, including Los Angeles’, have even started using body cams for real-time facial recognition of people officers are talking to on the street or during traffic stops. But the system is far from flawless. One in seven of the FBI’s searches identifies an innocent party, even when the actual culprit is in the database. And facial recognition has always been less reliable for people with darker skin, because of the way light reflects off it, who are already arrested in disproportionately high numbers.

What are the other risks with facial recognition technology?
The biggest danger is that authoritarian governments will use the technology to surveil and control their populations. Stanford University researchers made an algorithm that guessed someone’s sexual orientation from a picture of their face with 81 percent accuracy; humans managed only 61 percent. In countries where homosexuality is illegal, that could be a dangerous weapon. FindFace, a Russian app, can identify strangers by comparing their photo to more than 200 million social media profile pictures, and it’s been used to harass people.

Is there any regulation?
European regulators have proposed that all biometric data, including “faceprints,” belong to their owner and thus require consent to use. But U.S. lawmakers appear relatively unconcerned: Only Illinois and Texas have laws regulating facial recognition; of 52 police agencies that have acknowledged using the technology, only one obtained legislative approval. Facial recognition still isn’t as good as it is in the movies, with computers instantaneously identifying every individual in a huge crowd. But as of 2017, it’s not that far off, and it's definitely making spying much harder.

In late 2018, Israeli intelligence agents came up with clever tools to fool the facial recognition technology that threatened their espionage efforts. Facial recognition tech is so advanced that it can identify people even if they’re wearing fake beards and mustaches, so a firm founded by former military intelligence officer created software that can subtly distort an individual’s face in a digital photo, so that a computer won’t be able to match that image to a real person. They also have eyeglasses that use infrared light or flashes, as well as an infrared device that can be hidden under a baseball hat, umbrella, or wig and project dots of light onto the wearer’s face to fool facial recognition. The recognition/anti-recognition tech race is roaring ahead at full speed.