The ability of facial recognition technology to successfully establish an individual's identity has got 20 times better in the past four years. The success of facial recognition software capabilities increased by 2,000% between 2014 and 2018, according to a report by the National Institute of Standards and Technology (NIST).
The report measures the ability of biometric software to achieve a 'true positive': when an enrolled user is correctly matched to their profile. NIST looked at 127 software algorithms from 39 different developers, and tested its success in matching a person's photograph with a different photograph of the same individual stored on a database.
How facial recognition has achieved these unprecedented improvements
The research found that the software got 20 times better at successfully matching photographs in the four years of the study, pointing to a rapidly advancing marketplace for facial recognition-based biometric algorithms. Only 0.2% of searches failed in 2018, compared with 4% in 2014 (and 5% in 2010).
The success has been largely attributed to an industry uptake of convolutional neural networks, a type of machine learning commonly used to analyze imagery. The implication of this, according to a computer scientist at NIST, is that any organization not yet utilizing these convolutional neural networks should upgrade to this technology.
The two big tests of any authentication system are accuracy and security. The report from NIST confirms that accuracy in facial recognition software is improving; as a result, it becomes harder for hackers to gain access to these systems and individuals can be confident that their data is protected and secure.
Find out more about the technology behind facial recognition, and the security implications, on our microsite.