Chinese scientists at the Chinese Academy of Science (CAS) have recently developed a facial-recognition payment system, which has been dubbed “with a near-perfect accuracy”. The technology will be incorporated in an app that will be launched in 2015.
The new payment system is going to be available to both retailers and individual customers at the time of launch next year.
According to a report by the Science and Technology Daily on Sunday, the Chongqing-based CAS has said that it has set up the world’s largest Asian face database, currently containing more than 50 million records. It was created also with input from two other universities; the University of Illinois in the US as well as the National University of Singapore.
According to the director of the CAS’s Chongqing Green-Technology Research Institute, Zhou Xi, this unique data-collecting technology can gather facial information from 91 distinct angles, simultaneously! Researchers were able to also enhance its learning capabilities to function accurately and stably in a dynamic environment.
The system scored a high accuracy of 99.8% going by the Carnegie Mellon international standard, thus exceeding its predecessors’ best record of 97.6%.
But, is 99.8% good enough for financial transactions?
The new technology has not come without controversies, especially due to its being dubbed “near-perfect”. It has been criticized for having a “far too low” accuracy level for a financial-use technology.
99.8% accuracy means a 1 in 500 failure rate, which has been described as far from near-perfect accuracy when we are talking about money. In order to secure financial transactions, this figure needs to be a lot less than 1 in several hundred millions, something 99.999999’ish or even better if it has to be usable.
Considering how fast it can take to make 500 financial transactions using modern technology –even for a single individual – this technology is definitely not viable for commercial use yet.
So far, the facial recognition system has been tested at border control locations and attendance machines. Once users have installed the app on their mobile devices and linked it with their credit cards or bank accounts, it takes one second for the system to correctly identify the payer, according to Zhou.
To address the security concerns, Zhou says that the app’s special algorithm can easily detect whether a scanned face is actually the real one, instead of a held-up video or photo. The system has also been equipped with an online learning capability that automatically studies any changing characteristics of a subject to for instance, prevent weight changes from having any effect on the results.
As for the other accuracy and security concerns, who knows, maybe the concerned Chinese Scientists will have improved their app by the time of launching it into the market next year!