In short
How Ayonix face recognition is evaluated, what independent evidence exists, how liveness is tested against named attack instruments, and what governance a deployment needs. Every external claim links to its primary source; no accuracy percentage appears anywhere on this site.
Ayonix in NIST face recognition evaluations
What participation establishes, what it does not, and links to NIST’s own published records rather than a summary of them. Includes why NIST certification does not exist.
Accuracy and how to test it
Why no percentage appears on this site, the five variables that move any accuracy figure, the four error rates, and the method for establishing a number that describes your deployment.
Liveness and presentation attack detection
What a liveness claim must state before it means anything, why ISO/IEC 30107-3 issues no certificate, and the difference between presentation and injection attacks.
Privacy and governance
Fourteen controls a defensible deployment carries, what meaningful human review actually looks like, and the readiness checklist behind them.
House rules
The editorial rules these pages follow
Stated here so they can be checked against the pages, and used against us if we break one.
No accuracy percentage
For Ayonix or any other vendor. A figure without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced, so it is not a fact a buyer can use.
Sources over summaries
Where an external record exists, it is linked rather than summarised. A vendor summarising its own evaluation results is choosing which results to summarise.
Limits alongside claims
Every registered claim states what it does not establish, at the same visual weight as the claim. The limit is why the claim can be trusted.
No invented interfaces
No SDK function name, API endpoint or platform matrix appears unless it has been verified. Developers copy code from vendor pages, so this one matters more than it looks.
No certification that does not exist
NIST does not certify. ISO/IEC 30107-3 issues no certificate. Both phrases are on the banned list and the build refuses them.
Corrections in public
When text changes, the last-reviewed date changes with it. We would rather be described accurately than favourably, including by ourselves.
Ayonix algorithms have been submitted to NIST face recognition evaluations
- Source type
- Government or standards body
- Verified
- 2026-09-11 · Dr Sadi Vural, Chief Executive Officer
What this does not establish
NIST publishes comparative evaluations and does not endorse, certify, approve or accredit vendors. Participation means an algorithm was measured on data the submitter did not choose. It is not a statement about performance at any particular site.
1:1 verification, 1:N identification, watchlist matching, face tracking, and liveness and presentation-attack detection
- Source type
- Ayonix first-party statement
- Verified
- 2026-09-11 · Jan Mocary, Chief Technology Officer
What this does not establish
Presentation-attack resistance is specific to the attack instruments tested for. Liveness is available in supported capture workflows, not universally.
Frequently asked questions
What independent evidence exists about face recognition algorithms?
The strongest is the evaluation programme run by the US National Institute of Standards and Technology, which measures submitted algorithms on sequestered data under fixed conditions and publishes the results whatever they show. NIST does not endorse vendors and issues no certification, so a report is evidence about an algorithm rather than a recommendation of a supplier. Everything else in this market is measured by the party selling it.
Why does this site publish no accuracy percentage?
Because a percentage without its threshold, dataset, gallery size and demographic breakdown cannot be reproduced, and a buyer cannot act on a figure they cannot reproduce. The same algorithm can truthfully be called 99.9 per cent or 95 per cent accurate depending on those four variables. Accuracy for a specific site is established by a pilot on that site’s cameras and population.
What is the difference between a laboratory result and a site result?
They answer different questions and neither substitutes for the other. An evaluation says how an algorithm behaved on that data under those conditions. A pilot says what this installation will do with these cameras, this lighting and these people. Both belong in a procurement decision, and a vendor offering only the first has not answered the question you are actually asking.
Next step
Establish the figure for your site
Published evaluations describe algorithms. A pilot on your own cameras, with acceptance criteria agreed before it starts, describes your deployment. Only the second is something an operations team can plan against.