Active vs passive liveness detection
“Active or passive?” is the first question most teams ask about liveness, and it is usually framed as a security-versus-friction trade-off. That framing is now out of date. Here is how each actually works, where each wins, and why the right answer for most flows is not one or the other but both — passive-first, active when it counts.
Active liveness: challenge-response
Active liveness asks the user to do something — turn their head, follow a dot, blink, or perform a gesture — and checks that the response happens correctly and in real time. Its appeal is intuitive: a printed photo can’t turn its head. The cost is equally real: it adds seconds and instructions, it raises abandonment at exactly the moment you’re trying to onboard someone, and it assumes a level of vision, dexterity and comprehension not every user has.
Passive liveness: one selfie, no instructions
Passive liveness reaches a decision from a single captured selfie, with nothing asked of the user. It analyses signals an artefact struggles to reproduce — skin texture and micro-texture, depth and 3D cues, reflection and lighting consistency, and subtle involuntary motion. The experience is a normal selfie; the work happens invisibly. The trade is that it leans entirely on model quality, so it lives or dies by how well it’s trained and kept current.
Why the old trade-off broke
Active used to be the security-maximising choice. Two things changed. First, attackers learned to replay and inject responses — a recorded head-turn, or synthetic frames fed straight into the app, can satisfy a naive challenge. Second, passive models got good, reading cues that are genuinely hard to fake. So the modern question isn’t “which mode is safer” but “did these frames come from a real camera and a real person” — which is answered by combining face analysis with capture-path integrity and device signals, under either mode.
When to use each
Lead with passive
High-volume consumer onboarding where conversion and accessibility matter, and you want the common case to be one frictionless selfie.
Step up to active
When the passive signal is uncertain, the risk is high-value, or a regulator or policy expects an explicit challenge for a subset of cases.
Frequently asked questions
- Is active or passive liveness more secure?
- Neither is strictly more secure any more. Active challenges once had the edge, but recorded and injected responses can replay a head turn, while modern passive models read cues an artefact struggles to fake. The strongest posture is layered: passive analysis plus, where needed, an active step-up, both backed by capture-path integrity so injected frames are caught regardless of which mode ran.
- Which gives a better user experience?
- Passive, clearly: the user takes one selfie and is done, with no instruction to follow. Active adds a few seconds and cognitive load, which raises drop-off and can exclude users who can’t easily perform the action. That is why Zanyara runs passive-first and only asks for an active challenge when the passive signal is uncertain.
- What about accessibility?
- Active challenges assume a user can see and follow a prompt and physically perform a movement — a barrier for some people with visual, motor or cognitive impairments. Passive liveness asks nothing of the user, so a passive-first flow is the more inclusive default, with human review as the fallback rather than a harder challenge.
- Does Zanyara make me choose?
- No. Zanyara ships both and orchestrates them for you: passive on every check, an active step-up when confidence is low, and human review for the residual. You get passive’s conversion with active’s assurance where it matters.