NTN · 2026-11-12
Direct-to-Device at Scale: Identity Implications of Universal Coverage
By J. W. Bouckaert
The assumption under geo-based risk signals
Many risk rules treat a phone's network position as a rough location. A login that arrives through a cell in one country, minutes after one in another, gets flagged. A transaction from a network area the customer has never used gets a step-up. Those rules work because a terrestrial cell is small and sits in one country.
Direct-to-device satellite service changes the size of the cell. 3GPP's overview of non-terrestrial networks states the scale plainly: "With NTN it is possible to deploy very large cells over large portions of a continent (possibly covering different countries), with the different core networks for the various countries connected to the same NTN RAN where the serving cell information may not be granular enough to know a UE's location." For LEO constellations it adds that the satellites are "moving with high velocity which leads to moving cell patterns", and that "the moving cells may also cross country borders or cover areas with no territorial claims."
When the fallback path is a cell of that size, the serving cell stops being a usable location signal for anything finer than a region.
Location in the standards
3GPP did not leave location to the cell. Three mechanisms in the NTN work bear on identity.
| Mechanism | Release | What it does | Source on the 3GPP overview |
|---|---|---|---|
| Coarse UE location report | Rel-17 | Upon network request, after AS security is established, the UE "should report its coarse UE location information (most significant bits of the GNSS coordinates, ensuring an accuracy in the order of 2 km)" | NR NTN, RAN2 and RAN3 |
| Country-aware core selection | Rel-17 | When the gNB "detects that the UE is in a different country to that served by the serving AMF", it should hand over to an appropriate AMF or start a context release | RAN3 AMF (re-)selection |
| Network-verified UE location | Rel-18 | TR 38.882 "concluded on the need for a network based solution which aims at verifying the reported UE location information (within 5-10km, like the diameter of a terrestrial macro cell)"; Rel-18 NR_NTN_enh defines UE and gNB Rx/Tx time measurements to support it | RAN study FS_NR_NTN_netw_verif_UE_loc and WI NR_NTN_enh |
The third row is the one to read twice. The study's stated drivers are "public warning systems, emergency call support, legal interception": regulatory needs that cannot accept a location the handset simply asserts. 3GPP's answer was to verify the reported location from the network side, to roughly the precision of a terrestrial macro cell. The target is 5 to 10 km, coarser than the 2 km report the UE sends.
The identity implications
Three consequences follow for anyone who scores risk from network position.
A country change is not evidence of travel. If one satellite cell covers several countries and the core network is chosen by the UE's reported position, the country a session appears to come from can change without the device moving. Rules that treat a cross-border hop inside minutes as impossible travel will fire on satellite users who stood still.
The handset's own location claim is the weakest input. The standards treat the GNSS-derived report as something to verify. A risk system should treat any device-asserted location the same way, and weight a network-verified location above it once operators expose one.
Coverage gaps stop being a signal. On terrestrial networks, a device that is unreachable is often somewhere without coverage, and that can be informative. With a satellite fallback, reachability tells you less about where the device is and more about which path it took.
None of these remove geography from risk scoring. They move it from a precise signal to a coarse one with a known error bar, which is how a rule should treat it anyway.
PasskeyBridge and location today
PasskeyBridge stores IP addresses only as keyed digests, HMAC-SHA-256 under a server-held pepper, so a stored address cannot be mapped to a place. The insider-threat analytics that scan a tenant's audit log include a detector named impossible travel, and it does not measure distance: it flags two logins from different IP digests less than five minutes apart. A device that changes network path between terrestrial and satellite access can present a new address, and a rule like that reads the change as a second source. That is the correct reading of what the rule can see, and it is why the detector records an anomaly and takes no action on its own.
Carrier lookups report SIM swap, porting and reachability where the provider answers. The roaming and location-verification insights some carrier APIs list in preview are not requested. When operators expose a network-verified location through an API with a defined error bound, it becomes a signal we can consume; until then, nothing in the product infers location from the network path.
For the handoff between terrestrial towers and satellites and its effect on SIM attestation, see our earlier post on Starlink Direct-to-Cell. For what 3GPP's next release changes in satellite access security, see 3GPP Release 19 NTN security.
Questions for the annual review
- When will operators expose network-verified UE location to relying parties, and with what error bound and consent model?
- Will country-aware core selection give relying parties a usable "serving country" field, and how often will it change for a stationary device?
- How should impossible-travel rules be re-tuned for users whose fallback path is a continental cell?
Frequently Asked Questions
Why are satellite cells a problem for geo-based risk rules?
3GPP notes that NTN cells can cover large portions of a continent, possibly several countries, and that LEO cells move and can cross borders. A serving cell that large is not a usable location for anything finer than a region.
What is network-verified UE location in 3GPP?
A Release 18 capability, from study TR 38.882 and work item NR_NTN_enh, that verifies the location a device reports using UE and gNB timing measurements, to within roughly 5 to 10 km, for needs such as emergency calls and lawful interception.
Does PasskeyBridge use IP geolocation?
No. IP addresses are stored only as keyed HMAC-SHA-256 digests, which cannot be mapped to a place. The impossible-travel detector in its audit-log analytics flags two logins from different IP digests less than five minutes apart and records an anomaly.