Satellite connectivity has moved from a specialist tool to something ordinary. It is in aircraft cabins, on ships, on remote worksites, in emergency response, and increasingly in phones through direct-to-cell services. It is also the transport people reach for specifically when the terrestrial network is unavailable or untrusted, which makes its actual properties worth knowing rather than assuming.
The footprint is the threat model
A geostationary satellite sits at 35,786 kilometres above the equator. From there its transmit beam covers a large fraction of a hemisphere, and every earth station under that beam receives the same signal whether or not the traffic is addressed to it. Reception requires nothing from the operator: no account, no network access, no route to the target, no touching of any cable.
That inverts the usual assumptions about interception. Terrestrial interception implies a position on the path, and taking that position leaves traces. Satellite reception is entirely passive, leaves no trace at either endpoint, and cannot be detected by anyone. It is also cheap, because the same hardware people buy for satellite television will tune the relevant bands.
Low earth orbit constellations narrow this considerably, since a beam from a few hundred kilometres up covers a much smaller area and moves. That changes how close an observer needs to be, and it does not remove the observer.
The satellite phone ciphers were broken in public
Handheld satellite phones do not use GSM or LTE. They use the GMR family of standards, in two main variants matched to different constellations. Both variants shipped with proprietary stream ciphers whose designs were never published.
In 2012, researchers at Ruhr University Bochum reverse-engineered both ciphers out of publicly available handset firmware and presented the results at the IEEE Symposium on Security and Privacy. One turned out to be closely related to A5/2, the deliberately weakened GSM cipher that had already been broken years earlier. The other fell to its own analysis. Later work reduced the attack on the second cipher to something that runs in real time on ordinary hardware.
Both designs held up as long as they did because nobody outside the vendor could look at them, which is the failure mode Kerckhoffs described in 1883. When the firmware became available, both fell within a single research effort.
What was found simply listening
In 2020 a researcher at Oxford presented work at Black Hat that took a different approach: rather than attacking a cipher, point a dish at satellites and see what is not encrypted at all. Using roughly three hundred dollars of off-the-shelf home television equipment, the work sampled traffic from eighteen satellites and found substantial volumes of it readable.
The traffic included passenger sessions from in-flight connectivity, communications from ships at sea, and internal traffic from corporate and ISP networks using satellite backhaul. Some of it carried credentials and session tokens, which is the difference between passive observation and the ability to act.
A larger scan of geostationary downlinks published in 2025 found the same category of problem still widespread, including unencrypted backhaul carrying mobile network traffic. No single operator's carelessness accounts for that. The assumption that nobody is listening up there has been tested repeatedly by people with cheap equipment, and it has not held.
Satellite bandwidth is expensive and the round trip to geostationary orbit adds around half a second. Encryption and its handshakes cost both. A lot of deployed VSAT equipment was specified when that tradeoff was made in favour of throughput, and the equipment on a ship or a remote site has a service life measured in decades.
What modern constellations actually encrypt
Newer services are considerably better on the link layer. The commercial low earth orbit broadband providers encrypt the connection between the user terminal, the satellite and the ground gateway, so the passive dish attack that works against legacy VSAT does not apply in the same way.
What that encryption is for is worth stating precisely: it protects the segment between your dish and the operator's ground station. At that ground station the traffic is decrypted and handed to the internet, exactly as it would be at any ISP. The provider occupies the position an ISP occupies, sees the destinations you connect to, and is subject to the legal process of the jurisdictions it operates in.
| Segment | Typically encrypted? | Who holds the key |
|---|---|---|
| Legacy VSAT, terminal to satellite | Frequently not | Nobody |
| Satellite phone voice, older standards | Yes, with broken ciphers | The operator |
| Modern LEO broadband, terminal to gateway | Yes | The operator |
| Gateway onward to the internet | Whatever your application does | You and the site you reach |
| Application content under end-to-end encryption | Yes, across all of the above | You and your correspondent |
The last row is the one that does the work. An end-to-end encrypted message crossing an unencrypted VSAT link is still unreadable to a dish pointed at the beam. The observer gets the fact of the connection, its timing and its volume, which is traffic analysis rather than content.
Direct-to-cell inherits the cellular model
Services that let an unmodified phone connect to a satellite work by putting a cellular base station in orbit. Your phone uses its normal SIM identity and the operator's normal core network. Everything that was true of your mobile connection remains true: the same subscriber identifiers, the same operator records, the same lawful intercept obligations that apply to terrestrial cellular.
There is one incidental change in your favour. Locating a phone by which cell it is attached to is much coarser when the cell is a satellite footprint rather than a sector antenna on a rooftop. That is a side effect of geometry, not a privacy feature, and it does not survive the phone also reporting GPS.
The uplink is a transmitter
Everything above concerns what a listener can read. A separate property applies to the terminal itself: a satellite phone or a VSAT dish is a directional radio transmitter with a distinctive signature, and transmitters can be located by the people who care to. This is well understood by anyone who has worked in a conflict zone, and it is the reason satellite communication discipline in those environments is about when and from where you transmit rather than about what the traffic contains.
Related work has shown that the downlink signals from broadband constellations can be used for positioning by third parties with no cooperation from the operator, by treating the constellation as a navigation source. That research is aimed at navigation resilience rather than at surveillance, and it is a reminder that a dense constellation of known transmitters is a public reference system whether or not it was meant to be one.
A satellite link reaches you by covering everyone else at the same time. Treat it the way you would treat any network you do not control.
The practical version
- Assume the link is public and encrypt above it. This is the same conclusion as untrusted Wi-Fi, with a larger radius.
- Know what a VPN buys here. It encrypts the segment your provider would otherwise read and moves that visibility to the VPN operator. It does nothing for the endpoints. The limits are the same limits as on the ground.
- Check for legacy equipment specifically. On a vessel or a remote site, the satellite terminal is often the oldest networked device present, and its defaults were set under a different threat model.
- Separate the content question from the presence question. Encryption addresses the first. Only transmission discipline addresses the second.
Haven encrypts message contents on the device before anything reaches a network, so the transport underneath can be a fibre, a hotel access point or a satellite beam covering a continent without changing what an observer of that transport gets. Link encryption and end-to-end encryption both get described as encryption, and the table above is where they come apart.