18. Device and Platform Matrix¶

Who should read this page¶

This page is mainly for SDK teams, mobile engineers, web engineers, QA teams, and product owners deciding which channels and device classes the liveness system must support.

Why this page exists¶

Liveness performance is not only a model issue.

It is also a platform issue.

The same system can behave very differently across:

Android app
iOS app
mobile web
desktop web
kiosk or assisted devices

That is why a device and platform matrix should be part of planning, evaluation, and rollout.

A simple platform view¶

Platform	Typical strengths	Typical risks
Android app	good camera access, device signals available	wide device diversity
iOS app	more controlled environment, strong capture consistency	fewer deep device controls in some cases
mobile web	broad reach, lower install friction	browser and camera limitations
desktop web	easy onboarding access	weaker camera quality and higher replay risk
kiosk / assisted	controlled hardware possible	operational complexity

Main dimensions to track¶

A useful platform matrix should include:

platform
OS version
app version or browser family
device class
front camera quality
supported capture type
expected latency range
known limitations
policy differences

Example compatibility matrix¶

Channel	Capture style	Common limitations	Operational note
Android app	image or short video	huge hardware spread	evaluate low-end devices carefully
iOS app	image or short video	device-integrity signals may differ from Android	usually stable but still segment by model
mobile web	image or short video depending on browser	browser permission and camera quality variance	strong fallback and UX guidance matter
desktop web	webcam video or image	low-quality webcams, screen replay exposure	stricter policy may be needed

Low-end vs high-end device behavior¶

Device class matters because it affects:

camera noise
exposure control
focus stability
frame rate
CPU / GPU speed
latency and battery behavior

A model that is fine on flagship phones can behave much worse on low-end devices.

Browser-specific concerns¶

Web deployments should explicitly consider:

media-device permissions
browser family and version
secure context requirements
webcam resolution and frame-rate limits
tab switching and session interruption
virtual camera exposure and spoof risk

Do not treat “web” as one uniform channel.

Policy may differ by platform¶

A realistic program may use:

different thresholds by channel
stronger retry guidance on web
stricter policy for desktop replay risk
active challenge on some channels but not others
reduced feature set where one platform lacks support

This is normal if it is documented and evaluated.

What to test for every platform¶

Test area	Examples
capture quality	blur, low light, backlight, face size
user flow	permission denial, retry guidance, recovery path
security	replay, virtual camera, injection where relevant
performance	latency, CPU use, memory use, crash rate
release stability	app version / browser version regression

Suggested rollout matrix¶

flowchart TB
    A[Lab validation] --> B[Internal beta]
    B --> C[Small guarded<br/>rollout]
    C --> D[Segment review<br/>by channel]
    D --> E[Wider rollout]

This is especially important when supporting mobile web or desktop web for the first time.

Common mistakes¶

Mistake	Why it hurts
testing only flagship devices	poor field readiness
treating web as one channel	browser-specific problems stay hidden
using one threshold everywhere	platform differences get ignored
not logging device and browser versions	debugging becomes slower
shipping to all channels at once	rollout risk increases

Final takeaway¶

A good liveness system should know where it is running.

Platform differences affect:

capture quality
attack exposure
latency
reliability
policy design

That is why device and platform thinking should be built into evaluation, release, and monitoring.

Need term help?¶

If any technical terms on this page feel dense, use Appendix A1 — Key Terms first and then jump to the relevant appendix page for deeper detail.