Accessibility in biometric verification: the gambler that most liveness solutions leave out.
Many biometric liveness verification solutions are designed for a very specific scenario: a new phone, a good camera, a stable connection, and a face that fits the standard the model recognizes most easily. A considerable portion of the betting base in Brazil falls outside this scenario, and it is precisely this group that feels the brunt of a poorly calibrated verification process.
When failing doesn't mean fraud.
When the liveness system doesn't account for these variations, the result is usually a rejection: a legitimate bettor who can't complete registration, not because there's fraud, but because the flow wasn't designed for their device or context. For the operator, this is indistinguishable from successfully blocking fraud, when in reality it's legitimate revenue lost at the gateway.
What is left out of the ideal scenario?
Devices with cameras below 720p, older operating systems, unstable mobile networks outside major urban centers, and faces that deviate from the standard that most models are trained to recognize—cases of vitiligo, albinism, scars, facial paralysis, or Down syndrome—are examples of real variations that a rigid liveness flow tends to mistakenly reject.
Text-only instructions also exclude part of the audience. A flow that relies on reading to guide the user leaves out those who have difficulty reading or use screen readers, and an app that only works well with precise touch on the screen excludes those with motor limitations.
Why this matters for compliance, not just for experience.
Rejecting a legitimate user due to technical incompatibility has the same practical effect as a security breach: the customer abandons the registration and doesn't try again. The difference is that, in this case, the problem isn't with the fraudster, but with the solution's design. For an operator already working with a tight conversion margin in a competitive market, each technical rejection is revenue that goes directly to the competitor who managed to validate that same user.
What does an accessible verification actually include?
Experts point to several minimum features for a liveness workflow that doesn't penalize legitimate user variation: automatic capture that triggers on its own when the face is framed, without relying on touch precision; spoken instructions, in addition to written ones, guiding the capture; compatibility with screen readers like TalkBack and VoiceOver; tolerance for simpler cameras and unstable connections, including 3G networks; and a facial recognition model specifically validated against atypical faces, not just against the average pattern used in most training databases.
The pain that Legitimuz understood
Since the company's inception, accessibility has always been part of product planning and updates.
Since adopting technologies that take Brazilian faces into account, it also has technological elements that allow for good captures with low-resolution or even broken cameras.
Legitimuz is a Brazilian idtech company specializing in identity verification, biometrics, and fraud prevention, with a significant presence in the iGaming, fintech, and other regulated industries. The company holds ISO/IEC 27001, iBeta PAD Levels 1 and 2, BixeLab PAD Levels 1, 2, and 3, BixeLab IAD, and ISO/IEC 30107-3 certifications.
