3DiVi urges banks to track real-world facial security

3DiVi has set out a case for stronger, ongoing performance measurement in facial biometric security used by digital banking services, with an emphasis on metrics that show how systems behave across devices, networks and attack patterns.

The company develops computer vision software and sells biometric and anti-fraud tools that banks and other digital service providers use for remote identity checks and authentication.

Beyond features

3DiVi said financial institutions often describe their biometric security in terms of functions such as face detection, liveness checks and anti-spoofing. It said that approach leaves a gap when services move from testing into production environments.

In live deployments, banks face variation in camera quality, handset models, lighting conditions and network stability. Fraud tactics also change over time. 3DiVi said those factors can shift error rates and user drop-off in ways that product feature lists do not show.

In facial biometrics, it's easy to fall into the trap of hope, said Mikhaylo Pavlyuk, Digital Identity Consultant, 3DiVi.

3DiVi said the core question for banking security teams is how likely the system is to fail in practice and whether that risk changes. It said teams need to quantify false acceptance and false rejection, and then monitor those rates across channels and customer groups.

"But once you go live-different devices, varying camera quality, unstable networks, evolving attack patterns-those possibilities no longer answer the key question: What is the probability of failure, and how does it change over time?," said Pavlyuk.

Operational metrics

3DiVi said discussions about biometric security often focus on what technology can do rather than what it delivers. It said teams should treat performance metrics as operational controls, rather than reporting outputs.

"This is why possibilities can be dangerous-they steal focus. We end up talking about what a system can do, rather than how much it can be trusted in the real world," said Pavlyuk.

The company pointed to common claims that it said require measured outcomes. It cited liveness detection rates, device-specific degradation in performance, and the extent to which anti-fraud systems reduce losses rather than raise alert volumes. It also flagged the resilience of face matching under low light, low resolution and video compression.

Product framing

3DiVi used its Biometric Anti-Fraud product, 3DiVi BAF, as an example of how it structures controls around measurable results. It said the product combines face biometric verification and identification with defences against biometric attacks such as liveness checks, anti-spoofing and synthetic detection. It also includes session-level anti-fraud using behavioural and environmental signals.

3DiVi said it adds a layer focused on performance metrics. It said this layer measures outcomes and risk in real-world conditions. It said the metrics align with NIST SP 800-63-4.

In biometric systems, the real challenge isn't what the technology can do but understanding how well it actually performs in the real world, added Pavlyuk.

What to track

3DiVi listed several operational measures it said banking teams can use to manage authentication risk and customer experience. These include passRate and failRate, which show the share of successful and failed attempts. It also highlighted completion time as a measure of how long users take to complete authentication.

The company also cited suspectedFraud, which represents the percentage of attempts flagged as risky. It identified abandonmentRate as a measure of how many users drop out during the process. It also highlighted fraudProofing and fraudAuthentication as indicators of fraud observed at different stages. It listed authenticationFailures as the average number of failed attempts per user.

3DiVi said teams can use these measures as a live dashboard, with attention on where users drop out and where risk signals rise. It said banks can then adjust thresholds and controls based on observed outcomes rather than assumptions.

Calibration and drift

3DiVi set out an approach it said can turn performance measurement into routine decision-making. It said teams should use metrics to calibrate thresholds and rules. It also said organisations should watch for drift, such as rising abandonment that could reflect user experience problems, device shifts or network issues, and rising suspected fraud that could indicate changes in attacks or audience mix.

The company also said session-level anti-fraud adds context about how trustworthy a given attempt appears at the time, alongside a biometric match decision. It said teams should assess how that context affects measured outcomes.

3DiVi said performance measurement can shape how banks balance fraud risk against customer friction, particularly as digital onboarding and account access continue to shift towards remote and mobile channels.