Perhaps what is looming is an invasion of privacy of such magnitude that not even the most fantastic Orwell could have imagined it. We have created computer programs that recognize faces, systems capable of identifying a specific person in a crowd. There is also already software that guarantees that the person who is registering on a website , withdrawing money from an ATM or starting a car is who he claims to be.
Although still supervised by humans, facial recognition machines already operate at various airports in the United States to verify the identity of travelers. Its operation –as an automatic teller machine– is very simple, and the screen will tell the user the steps to follow. First, pass the passport chip through the reader. Then check that the personal data shown is correct and, finally, pose in front of the camera. Click.
In a few seconds, the system compares the parameters of the photo stored in the passport chip with those of the image it has just captured and decides if there is a match between the two. Finally, print a receipt that the user must deliver at the next police checkpoint. Today the system is still in the testing phase, so going through the machine is one more procedure within the cumbersome security circuit of any American airport. However, in short, this technology will streamline processes and improve the reliability of identifications. It will be practically impossible to travel with false documentation.
Identity theft is one of the biggest headaches for the police, and not just at airports. Any process that requires identification through a computer system is liable to be desecrated. Passwords, usernames and other codes fall short of the permanent threat of hackers.
One of the most concerned sectors in this regard is the banking sector. 74% of European banks have increased investment in security and resilience of their information technology systems in the last year, according to the European Banking Authority (EBA) risk survey from the first quarter of 2015. And almost 60% of financial institutions have plans to further strengthen management of these threats in the short term . An effort that might not be enough: “The risk of cyberattacks is growing, the threats are becoming more complex and intense,” the European Central Bank (ECB) warns in its Financial Stability Report.
The Spanish company FacePhi, based in Alicante, works in this swampy scenario and has been investing in the development of facial recognition software for eight years. “The ideal thing to cover the security triangle is something I have, like a mobile phone or a credit card ; something I know by heart, like a password, and something that I am, and that’s where biometrics comes in “, explains Javier Mira, co-founder and vice president of FacePhi. This company has just signed its first major international contract with BBVA Previsión AFP, the Bolivian subsidiary of the Spanish bank, for the implementation of its software among clients.
It will first be aimed at pensioners, in order to avoid cases of impersonation or management of the pension once the beneficiary has died. Then it will be extended to all users of your online banking. “Thanks to our technology, a camera like that of any computer or mobile – Mira clarifies – detects if there is a face and, with some algorithms, generates a facial pattern of it, something like a giant password . The pattern is compared with the entity’s database , which deduces with very little margin of error and in a matter of seconds whether or not it is the correct person ”.
FacePhi’s patented software incorporates a retraining system, so that the device is able to identify a person even if some features of their face change due to causes such as age, weight, etc. “If I know you today and I don’t see you for another year, then your face will probably sound familiar to me, but I may not know how to locate you. However, your family and friends recognize you even if you are on your side, wearing glasses or growing a beard. That’s because their heads have registered many images of you, and they identify you very effectively. Our technology works the same: the more times you register, the easier it will identify you, even if there are visible changes ”, explains the FacePhi manager.
If a face, vectorized by software, becomes a code, then can it be hacked ? “In technology everything is hackable. 100% security does not exist. The facial pattern incorporates metadata with, for example, time keys, which means that, if someone is able to steal the code, they would have seconds to act; later that pattern would no longer be of any use . Nor can it be reverse engineered: from a stolen facial pattern it is not possible to reconstruct a face ”.
Academic research on facial recognition in Spain has its spearhead in the Facial Recognition & Artificial Vision Group of the Rey Juan Carlos University of Madrid (FRAV), led by Dr. Enrique Cabello. The FRAV has spent years developing systems with its own resources or through European programs, such as the ABC Gates For Europe (ABC4EU) project , which seeks to automate borders or road safety, such as the detection of driving behaviors.
“Now we are trying to advance, for example, in what is called spoofing detection or antispoofing . Imagine that you have a facial recognition system that works well, but you put a 3D mask with the mold of your face on and it does not realize it. Or that you go to the ATM, get money with your face and a hacker steals your facial data. You can change the card or pin, your face cannot. We design systems with biometric data as a cryptographic key so that only you can enter with your face, ”says Cabello.
Security is one of the key concepts – although not the only one, as we will see later – to understand why companies and administrations are betting on biometrics. One of the fields where more fruits are expected to be collected is the search for criminals. In this sense, the FBI has just shown the way, after arresting a pedophile who had been in search and capture for nineteen years. Lynn Cozart, now 63, sexually harassed three boys in Pennsylvania in 1996, but fled two months before her trial. Cozart was able to disappear from the map for almost two decades, but finally, last July, the FBI was able to get hold of him thanks to the Next Generation Identification (NGI) system.
This software compares crime photos of criminals against image databases of all kinds, such as photographs from state agencies, police videos, and security cameras. In this case, the system made a match between Cozart’s photo taken in 1996 and that of a driver’s license issued in Arkansas in the name of one David Stone. It was him. The NGI system contains nearly 125 million fingerprints and 24 million mugshots, and is capable of comparing tattoos, scars, or voices . But the use of images captured not only by the police or the state, but also by private companies, has reopened the privacy debate. Many see in this massive capture of images a threat to anonymity.
If the technique to compare the biometrics of two still images is a challenge practically overcome, the identification of a certain person in a crowd using IP cameras is another of the paths that facial recognition is traveling at full speed . This is the area where the Spanish company Herta Security, headquartered in Barcelona, develops its business, whose name jumped to the media last April when its software was applied in the security operation for the delivery of the Golden Globes, in The Angels. The idea, in this case, was to prevent star stalkers and party crashers from spoiling the party.
To this end, Herta Security had registered the faces of some forty people who could cause problems in this type of ceremony. It was about monitoring the crowd gathered around the red carpet to sift through possible suspects . The software performs a scan in real time and detects the faces, extracts an image of each one, parameterizes it with algorithms and converts it into vectors, which are compared with the patterns in the database. If there is a match – after 50% agreement – the system alerts the operators, who, of course, must personally corroborate the suspicions.
To interpret feelings, programs again need patterns. In this sense, the companies involved are based on the work of the American psychologist Paul Ekman, now in his eighties, who has spent more than four decades cataloging microexpressions to study emotions. Its Facial Action Coding System collects more than 5,000 muscle movements and identifies their emotional significance.
From the premises of Ekman, the mentioned companies are building gigantic databases that are already part of their business tools. For example, Affectiva claims to have measured 7,000 million emotional reactions from 2.4 million videos made in 80 countries. Because we are not all the same: “What adds more complexity are cultural variations: for example, that the Spanish laugh more openly than the Japanese,” says Cabello. Affectiva claims to have learned during the elaboration of her catalog of gestures and emotions that women smile more than men , or that South Africans are the most expressive in the world.
