Identy.io’s latest results in NIST FRIF TE E1N demonstrate what happens when accuracy, speed, and scalability are maintained simultaneously, precisely where biometric identification is most demanding.
An ABIS exists to answer one question: Who is this person? It does not start with a claimed identity to verify. Instead, it searches for a fingerprint among millions of records to identify the person it belongs to. This is known as 1 identification.
This is also why NIST results carry so much weight in the industry. NIST (National Institute of Standards and Technology) is the U.S. government’s standards institute and independently evaluates biometric technology. Its FRIF TE E1N evaluation measures precisely this capability: how effectively an algorithm can identify an individual within galleries containing millions of records, comparable in size to those used by national identity systems.
Identy.io's Results
-
99.99% correct identification rate using ten fingerprints against a gallery of 3 million subjects, with the false positive identification rate limited to 0.1% (FPIR ≤ 0.001).
-
2.9-second median search time per plain fingerprint search in a gallery of 3 million subjects.
-
Consistent performance across index fingers, slap impressions, ten-finger searches, plain-to-rolled comparisons, and all three test categories (A, B, and C).
Accuracy, Speed, and Scale: All Three or None
Each of these metrics matters because, in 1 identification, no single variable is sufficient on its own.
Accuracy without speed limits operational efficiency. An engine that delivers accurate results but takes minutes to do so forces organizations to overprovision infrastructure or accept queues, backlogs, and response times that end customers will not tolerate.
Speed without accuracy creates risk. Every false match means a person has been incorrectly identified and an examiner must review the wrong candidates. That is why NIST measures accuracy under a false positive identification rate (FPIR) threshold: without that limit, accuracy figures can be artificially inflated.
And both must remain consistent as databases grow. An algorithm that performs well with thousands of records will not necessarily deliver the same results with millions. The galleries used in this evaluation, containing up to 5 million subjects, are specifically designed to test that capability.
What This Means for Those Building on This Technology
For system integrators and technology providers, this translates into a tangible competitive advantage: the ability to offer end customers timely, accurate, and fast identification, regardless of database size. They can support their proposals with independently validated results, which often serve as a benchmark in competitive procurement processes. They can plan and size projects with confidence because every second saved per search reduces the additional infrastructure required to maintain the same level of service. And they can address their customers’ most important concern with compelling evidence: the engine has already demonstrated how it performs at scale.
NIST does not sell algorithms or represent vendors. It evaluates all participants under the same protocol, using data that participants do not control. Its evaluation does not replace testing within each project’s specific environment, but it provides one of the strongest foundations for comparison before reaching that stage.
What This Looks Like in Practice, by System Type
-
National Identity and Civil Registration: Every new enrollment requires a search against the entire population database to rule out duplicate identities. An accurate, high-speed engine enables systems to issue identity documents without bottlenecks or duplicate records.
-
Border Control and Immigration: Response times are measured in airport queues. Searches completed in seconds allow systems to check watchlists and previous records without slowing passenger flow, while controlling false matches helps prevent innocent travelers from being incorrectly flagged.
-
Law Enforcement and Public Safety: When a fingerprint is the primary lead, a correct identification can guide an investigation, while an incorrect one can send it in the wrong direction. In this context, limiting false matches protects individuals and preserves the system’s credibility.
-
Large-Scale Deduplication: Population registries, social programs, and customer databases need to ensure that each individual has only one record. Performing millions of comparisons accurately makes it possible to clean and deduplicate databases containing millions of records without sacrificing response times or accuracy.
Results That Improve with Every Evaluation
Between Identy.io’s two most recent NIST evaluations, the average identification error rate decreased from 0.117% to 0.095%. That represents a 19% reduction, with no performance regression in any category. For those integrating the algorithm, this means an identification engine that continues to improve with every version.
Among millions of records, your fingerprint is unique. And that is what these NIST results confirm.
HumanKey ID
Only you hold the key to your identity.
Considering an ABIS solution? (Request a Demo)
