What is NIST and what does it do?
NIST (National Institute of Standards and Technology) is a United States government agency charged with developing and promoting standards, measurements and technologies to improve the nation’s security and economic competitiveness. NIST works in areas such as information technology, cybersecurity, metrology, engineering, manufacturing and materials science. It is also responsible for creating and maintaining standards and guidelines in areas such as cryptography, information security and privacy. NIST is a very important organisation in promoting innovation and technological development in the United States.
NIST in the field of identity verification and biometrics
For more than 60 years, NIST has been dedicated to measuring and evaluating the performance of different biometric systems, including fingerprint, face, voice, iris and palm. Through these evaluations, they have gained unique insights into the advantages and disadvantages of each biometric, as well as the criteria that technologies must meet to be used in mass situations, such as identity verification using facial biometrics.
NIST has extensive experience in the field of biometrics, enabling it to conduct accurate and comprehensive assessments. In addition, it has access to unique US federal resources, which allow it to create assessment scenarios with real data. These datasets are very large, allowing them to measure the performance of systems with high accuracy, even detecting errors as low as one in a million.
Facial Recognition Vendor Test (FRVT)
It is the most respected benchmark when it comes to the evaluation of facial recognition algorithms (FRVTs) for verification (1:1) and identification (1:N). The evaluation uses very large sets of facial images to measure the performance of facial recognition algorithms developed in commercial and academic communities around the world.
Facial verification, called 1:1, refers to the confirmation or denial of an identity by comparing two biometric factors against each other (e.g. the image of a selfie against the image of an ID photo).
identification, also known as 1:N comparison, is the process of discovering an identity among a collection or set of people (e.g. set of face databases).
These evaluations are free of charge and open to any entity wishing to apply. Thus promoting the transparency and fairness of the results. Public results, making it the de facto reference for evaluating the quality of a facial recognition system.
Recently, they included a new assessment aimed at detecting presentation attacks (PAD). A presentation attack (PA), as defined in ISO/IEC 30107, is “the presentation of an artefact or human characteristics to a biometric capture subsystem with the intent to interfere with system policy”. A presentation attack is typically launched with the intent to imply (the user attempts to authenticate as a target identity) or evade (the user attempts to trick the biometric system into not recognising his or her true identity). Impersonation goals include attempting to gain a positive access privilege as another person, e.g. trying to unlock someone’s mobile phone or gain access to a facility. The goals of spoofing are often to conceal the true identity to evade recognition by, for example, a watch list, or to create a separate enrolment under a different name;
Biometric systems can potentially be attacked by an unknown number of presentation attack tools. Examples of known presentation attack tools include artificial “gummy” fingers, “replay” attacks in which the attacker holds a photo or video of someone’s face up to the camera, and iris photo and contact lens attacks.
Multiple databases are used for the evaluation.
At NIST, the databases used in facial biometric evaluations consist of sets of photographs of faces to measure the performance of systems in various tests. Among them, the most common are:
– Visa: photographs that comply with ISO/ICAO standards for facial photography, such as those used in passports or identity cards.
– Mugshot: photographs taken less rigorously than Visa photographs, usually by police during searches of criminals in prisons.
– Wild: photographs without any kind of control, which may come from news, newspapers, sporting events, etc.
– Border: photographs taken at border control facilities, such as airports.
Fuente de la imagen: informe FRVT 1:1 del NIST
Alice Biometrics, assessed and certified by NIST
At Alice, we are committed to transparency and regulation of biometric technologies, which is why we regularly submit our voice and facial biometrics engines for NIST evaluation.
We submit to NIST the biometric engines we deliver to real customers, avoiding adjustments that can be exploited by other players to get better results. And we rank among the best biometric engines in the world;
In June 2021, Alice is ranked in the top 10% of companies worldwide in facial biometrics assessments (FRVT 1:1). We have recently applied for the new FRVT PAD certification, being among the best solutions in this respect (results to be published by NIST shortly).