Dev.to Security 🔐 Cybersecurity 👁 0 📖 5 min read

OSNMA and Galileo HAS on a real autopilot: what you can switch on today, and how to verify it

We are UAV GNSS (uav-gnss.com), a Septentrio-powered receiver maker. Most integrators we talk to confuse two different jobs: proving the position is real and making the position accurate without a base station. Galileo's

We are UAV GNSS (uav-gnss.com), a Septentrio-powered receiver maker. Most integrators we talk to confuse two different jobs: proving the position is real and making the position accurate without a base station. Galileo's OSNMA and HAS are two separate answers to those two questions, and both are now switchable on a normal autopilot build without touching the flight controller. Here is what each one actually does, what it cannot do, and how to see it working in your logs.

Two problems that look like one

Spoofing is not a noisy signal — it is a clean one that lies. A spoofed receiver typically reports excellent C/N0, a healthy satellite count and a fixed-looking solution, because the receiver has no reason to doubt the data. Jamming is the opposite: the signal is bad and obvious.

Problem What it looks like What fixes it
Jamming / in-band interference Loud, visible: C/N0 collapse, fix loss, EKF resets Front-end resilience (filtering + adaptive notch), e.g. AIM+
Spoofing Quiet: good C/N0, plausible position, slowly drifting truth Cryptographic authentication of the navigation data (OSNMA)
Corrections unavailable Float or standalone, no cellular NTRIP link Free broadcast corrections (Galileo HAS), or a local base

If you only buy anti-jamming you are still open to a good spoofer. If you only buy authentication you still lose the solution in RF noise. They are separate line items.

OSNMA: the navigation message proves where it came from

OSNMA (Open Service Navigation Message Authentication) is Galileo's free, subscription-less authentication service. The constellation broadcasts a cryptographic signature chain alongside the navigation data, and a receiver with OSNMA support verifies it before trusting the satellites. A transmitter that did not come from the Galileo control segment cannot produce a valid signature, so a spoofer can imitate the ranging signal but not the authentication.

What it does not do: it authenticates Galileo only. GPS, GLONASS and BeiDou are not covered. On a multi-constellation receiver, that means a partial solution can still be authenticated while the rest is not, which is why you must look at which satellites are authenticated, not just whether the flag says "yes".

Septentrio's mosaic-X5 and mosaic-G5 have OSNMA integrated (see our page on anti-spoofing for drones for the receiver-side detail). In practice you configure it with a three-way mode, and the mode you pick is the whole design decision:

OSNMA mode Satellites used in the solution Host clock requirement Failure signature
off Everything the receiver tracks None Spoofing is invisible
loose All tracked sats, authenticated ones flagged None Solution survives, but you must read the flags to notice an attack
strict Only authenticated satellites Host must be NTP/PTP synced Under spoofing the solution degrades or drops — loudly

Strict is what most people actually want for security, but it costs you satellites and it needs a disciplined clock (the driver documentation puts the OSNMA firmware floor at GNSS firmware 4.12.1, with strict operation depending on host time sync). On a vehicle with no NTP, "loose + flag monitoring" is the honest configuration — and then you have to route those flags somewhere an operator sees.

If you run ROS, the official Septentrio driver publishes the authentication status as its own topic (/galauthstatus) next to /aimplusstatus and /rfstatus, so authentication state is observable in rqt or recorded in a bag like any other signal. That is the difference between "OSNMA is enabled" and "I can prove the solution was authentic at t=142 s".

Galileo HAS: decimetre accuracy with no base and no subscription

HAS (High Accuracy Service) attacks the correction problem instead. Galileo broadcasts free precise corrections two ways: signal-in-space on E6-B at 1278.75 MHz (448 bit/s per satellite, no internet needed), or over the internet via the IDD stream. The receiver applies them internally and outputs a better position over the same serial/Ethernet stream your autopilot already parses.

The official target is under 20 cm horizontal, and in our own measurements on a mosaic-G5 P3H module we have seen around 10 cm class results; the detail is in our HAS explainer. Three honest caveats:

  • It is not RTK. There is no base-to-rover integer fixing, so you do not get the pass-to-pass repeatability that row-to-row work depends on. HAS is a strong absolute-accuracy tier for mapping, corridor inspection and long-range flight where a base or a cellular NTRIP link is impractical.
  • It converges. Expect a warm-up of minutes after power-up, not seconds.
  • It needs open sky and a clean front end. Free corrections do not help if your receiver cannot track the E6-B signal or is being desensitised by an in-band interferer.

What changes on the autopilot? Nothing.

This is the part that surprises people. Both features live inside the receiver. The autopilot continues to receive the same SBF/NMEA stream at the same baud rate with the same GPS_TYPE it already uses — no new parameters, no new message definitions to parse. The difference shows up in the quality of what arrives, and in the status flags you choose to monitor.

What to verify, and where:

Check Where to look What "working" looks like
Galileo authentication Receiver status / /galauthstatus Authenticated satellite count roughly tracks the Galileo sats in view
Correction state Receiver PVT block (mode + correction age) Mode settles above standalone, correction age stays bounded
Solution integrity Autopilot EKF innovations / GPS status No step changes in position residual when switching sources
Spoofing response test Bench test with a simulator, strict mode on Solution refuses to follow the fake trajectory instead of gliding with it

That last row is the one worth scheduling before a customer demo. Enable strict mode, feed a simulator a trajectory that walks away from truth, and confirm the receiver's solution does not follow it. If you have never run that test, you do not know whether OSNMA is doing anything.

Where each one fits

  • Defence, BVLOS over people, port and critical-infrastructure work → OSNMA in strict mode, with the authentication status logged and alarmed. Here a wrong position is worse than no position.
  • Long-range survey, corridor mapping, remote operations with no base → HAS as the correction tier, with RTK where pass-to-pass repeatability matters.
  • Both, plus a hostile RF environment → neither OSNMA nor HAS stops a jammer. That still needs front-end resilience: the interference-mitigation chain in AIM+ is specified at roughly 40–60 dB versus about 25 dB for a typical consumer module, which is the difference between holding a fix next to a 4G site and watching the solution fall to float. The technology overview is on our AIM+ resilience page.

Wiring, parameter tables and the SBF stream checker are in our integration repo: github.com/uavgnss6-bot/septentrio-gnss-integration-guide.

Question for the room: if you run OSNMA today, are you in loose or strict mode — and what made you choose it? We are collecting real configurations, especially from builds without a disciplined clock on the host.

📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.