VU, one of five companies worldwide at iBeta Level 3: 2,700 attacks, none accepted

VU, one of five companies worldwide at iBeta Level 3: 2,700 attacks, none accepted

iBeta ran 2,700 presentation attacks against VU's facial liveness detection and none was accepted. VU's solution is compliant with iBeta Level 3 for ISO/IEC 30107-3.

November 18, 2026·6 min read·News/Announcement
Share:
VU Team
VU Team

CONTENTS
At a glance
  • The result. iBeta ran 2,700 presentation attacks against VU's Secure Onboarding Process, and none was accepted.
  • The level. Level 3 is the third of three: attacks with hyper-realistic masks, prepared by evaluators with at least 10 previous PAD tests.
  • The context. According to iBeta's public list, VU is one of five companies with a Level 3 letter and the first in Latin America.
  • The scope. The letter confirms compliance for a specific product version against attacks in front of the camera.

iBeta Quality Assurance confirmed that VU's facial liveness detection is compliant with Level 3 of its ISO/IEC 30107-3 testing program. It is the most demanding level of the lab's presentation attack detection (PAD) testing, and according to its public list only five companies in the world have reached it. According to that same list, VU is the first Latin American company to do so.

iBeta ran 2,700 attacks and none was accepted

From late August to October 29, 2026, iBeta evaluated VU's Secure Onboarding Process v2.28.0 (SDK v2.5.11), together with its server components. The lab is accredited by NIST/NVLAP to test against ISO/IEC 30107-3, and it signed the confirmation letter on October 29.

This is how the test was set up: six real people as reference subjects, three attack types built with high-preparation methods, and up to 56 hours of work per attack instrument. With all of that, iBeta ran 2,700 presentation attacks on two devices.

According to the letter, iBeta could not get the system to classify any of those attacks as a live person.

What sets Level 3 apart from the earlier levels

ISO/IEC 30107-3 defines the method, the metrics and how a presentation attack detection test is reported. The standard does not define the attack sophistication grades: iBeta's testing program does, ranking them by what it takes to build the instrument, in time, expertise and equipment. There are three.

Level 1 uses low-cost, easy-to-build artifacts. Level 2 adds more expensive instruments, on the order of a resin or latex mask. VU's facial liveness detection was already compliant with those two levels: we announced it in April 2026.

Level 3, which the lab added in 2025, raises the bar on three fronts:

  • The artifacts. Hyper-realistic face masks, with environments and accessories prepared for the attack.
  • Who attacks. Evaluators who spent at least 16 hours researching attacks on the modality and took part in at least 10 previous PAD tests.
  • The time. Up to seven days per attack type.

To be compliant, iBeta allows at most 5% of attacks to get through.

One of five companies on iBeta's public list, and the first in Latin America

iBeta publishes a list of the confirmation letters it has issued. The first Level 3 letter dates from January 2026. Since then, according to that iBeta public list, five companies have reached Level 3, and VU's letter is the fifth.

VU, headquartered in Argentina, is according to iBeta's public list the first Latin American company at this level.

What the letter confirms, and what it does not

The letter covers three specific things:

  • It is compliance, not certification. iBeta confirms that the test was run and reported according to ISO/IEC 30107-3, and that the result is at Level 3. The lab itself states that its letter is not a certification of the product.
  • It covers presentation attacks. These are the ones that happen in front of the camera: a mask, a photo, a screen. Injection attacks, which feed a fake video into the flow without going through the camera, are outside the scope of the standard, according to its ISO page. NIST defines them as a different attack and warns that presentation attack detection is not enough to cover them (NIST SP 800-63A-4).
  • It applies to one version. The letter names Secure Onboarding Process v2.28.0 and its SDK. It is the evaluation of that version, with its server components.

What changes for anyone evaluating a vendor

If you are comparing liveness detection vendors, a claim like "we detect deepfakes and masks" is not enough. What you can ask for is the lab's letter, the level it confirms and the version it covers.

VU's letter answers all three: Level 3, ISO/IEC 30107-3 and Secure Onboarding Process v2.28.0. To see how liveness detection fits into an onboarding flow, check out the Verify capability of VU ONE, for identity verification and onboarding.

shield
Check the letter and let's look at your flow. Download the letter from
iBeta's public list and compare it with what your vendors show you today. To see how VU's liveness detection performs in your own onboarding,
book a conversation.

Frequently asked questions

It is the most demanding of the three levels in iBeta's program for presentation attack detection testing with ISO/IEC 30107-3. It uses hyper-realistic masks, evaluators with experience in at least 10 previous tests and up to seven days per attack type.
IAPAR stands for Impostor Attack Presentation Accept Rate, the metric NIST requires for presentation attack detection in remote capture (NIST SP 800-63A-4). The iBeta letter reports an IAPAR of 0% for VU: none of the 2,700 attacks was accepted.
No. iBeta does not issue certifications: it issues confirmation letters. VU's letter confirms that its Secure Onboarding Process is compliant with iBeta Level 3 for ISO/IEC 30107-3.
No. According to its ISO page, ISO/IEC 30107-3 evaluates attacks that happen at the capture device, in front of the camera. ENISA describes video injection as a way to bypass the camera (ENISA report), and it is covered by other controls.

Want to stay up to date with the latest in digital identity?

Subscribe to VU's newsletter and receive use cases, industry news and articles on verification, authentication and fraud prevention.