Evaluation

On-device vs cloud facial recognition.

If you’re comparing approaches to photo consent screening, the biggest decision is where the matching runs. Here’s an honest side-by-side.

Same job, two very different data trails.

Where matching runs

On-device (FaceGate): Using the phone's local processing.
Cloud facial recognition: On the vendor's servers.

Where biometric data lives

On-device (FaceGate): The phone only.
Cloud facial recognition: Vendor infrastructure.

Breach surface

On-device (FaceGate): No central store to breach.
Cloud facial recognition: A central database of faces.

Cross-border transfer (APP 8)

On-device (FaceGate): None - nothing leaves the device.
Cloud facial recognition: Often, depending on hosting region.

Offline operation

On-device (FaceGate): Works with no connection.
Cloud facial recognition: Requires connectivity.

Procurement review

On-device (FaceGate): Narrower - no data-sharing agreement.
Cloud facial recognition: Broader - hosting, retention, DPAs.

Operator control

On-device (FaceGate): The operator holds everything.
Cloud facial recognition: Shared with the vendor.

Dimension	On-device (FaceGate)	Cloud facial recognition
Where matching runs	Using the phone's local processing.	On the vendor's servers.
Where biometric data lives	The phone only.	Vendor infrastructure.
Breach surface	No central store to breach.	A central database of faces.
Cross-border transfer (APP 8)	None - nothing leaves the device.	Often, depending on hosting region.
Offline operation	Works with no connection.	Requires connectivity.
Procurement review	Narrower - no data-sharing agreement.	Broader - hosting, retention, DPAs.
Operator control	The operator holds everything.	Shared with the vendor.

Cloud isn’t always the wrong choice.

Cloud platforms have real advantages: central administration across many sites, cross-device sync, and elastic scale. For some problems those matter more than anything else.

For screening children’s photos, the trade is specific - it means sending biometric data off the device to a system someone else operates. FaceGate takes the other side of that trade on purpose: the phone is the only computer involved. Which side fits depends on your risk posture, and that’s the decision this page is meant to make legible. The detail behind the on-device column is on the Privacy & Trust page.

FAQ

Evaluation questions, answered.

Is on-device matching less accurate than cloud?
Accuracy is a property of the model, not where it runs. FaceGate bundles the same production model on the device, so it doesn’t trade accuracy for privacy - see the model card on the Privacy & Trust page.
Doesn't cloud make multi-device use easier?
Yes - central administration and cross-device sync are genuine cloud strengths. FaceGate takes the opposite position on purpose: the consent list and the matching stay on the operating device rather than syncing through a server. Which matters more depends on how your team is set up.
What actually leaves the device with FaceGate?
No photos and no biometric data. The full data-flow table is on the Privacy & Trust page.
If there's no cloud, what happens when a device is lost?
This is the one place cloud has an edge - central recovery. FaceGate’s answer is that the device is never the system of record: your roster package - the file your organisation already holds - is. A replacement device re-imports the package and rebuilds the consent list from scratch, and your device-management platform handles remote-wipe of the lost one. The trade is deliberate: no central database of children’s faces to breach, and no biometric backup files to manage either. The detail is in the Privacy & Trust FAQ.

Comparing options for your organisation?

Tell us what you're evaluating - we'll give you a straight answer on whether FaceGate fits.

Book an intro call