Detection limitations FAQ
Unrecorded is built around honest limits. This FAQ explains what the app can notice, what it cannot know, and why every alert is a possible privacy risk indicator — not proof of recording.
Product truth: A match against known hardware patterns means a nearby signal may resemble smart glasses or a wearable recording device. It does not confirm that video or audio capture is happening.
Can Unrecorded prove someone is recording?
No. The app compares visible Bluetooth Low Energy (BLE) fields against a local catalogue of known patterns. When several weak indicators align, you may see a low, medium, or high privacy-risk level with plain-English reasons. That is situational awareness — not evidence of active recording.
Why is detection probabilistic?
- Hidden or randomised names — manufacturers and users can change what appears in advertisements.
- Spoofing — a device can advertise a misleading name.
- Noisy RSSI — signal strength is a rough proximity hint, not a measured distance.
- Weak address-prefix hints — partial address patterns are supporting evidence only, not identity proof.
- Platform filters — Android and iOS may limit which BLE advertisements your phone exposes to apps.
For technical background, see how smart glasses use BLE broadcasts and known smart-glasses BLE patterns.
Does scan data leave my device?
No. Unrecorded processes scans and scores risk on-device. Observations
are merged in memory inside a ScanSession during active
use. The app does not upload nearby Bluetooth results to
project-operated servers. Only your protection and notification
preferences are persisted — not a history of people or addresses
nearby.
When is scanning most reliable?
Scanning runs in short windows with rest periods to limit battery use. Keeping the app open in the foreground is the most reliable path when background protection is off.
On Android, optional background protection (off by default) runs a foreground service with a persistent notification where the OS allows. Battery saver, force-stop, and manufacturer policies may still pause scanning. The app explains honestly when protection was stopped by the system — it does not claim silent protection.
What do notifications mean?
| Notification | Meaning |
|---|---|
| Protection status (ongoing) | Protection is active while the app is running. Not proof that recording was detected. |
| Possible-risk alert | Nearby signals may match risk indicators. Tap for details. Not proof of recording. |
What can the app detect?
- BLE advertisements from nearby devices while scanning.
- Names, manufacturer hints, and cautious address-prefix patterns matched against a local open-source catalogue.
- Deterministic, explainable scoring — no machine-learning black box.
What can the app not detect?
- Whether recording is actually happening.
- Devices that do not use Bluetooth or hide advertising.
- Non-BLE cameras, microphones, or phones used as recorders.
- Every smart-glasses model that exists or will exist.
What about demo mode?
Demo mode uses a fake scanner with sample data for UI testing. It does not reflect live Bluetooth around you. Use it to understand alerts and copy — not as a field test of real-world detection.
Where to read more
- Privacy in public spaces — practical situational awareness
- Developer detection limitations (GitHub) — full engineering notes
- Privacy policy — data handling and permissions