NTN · 2026-03-14

Identity From Orbit: How LEO Satellite Signals Could Replace Ground-Based Entropy

By J. W. Bouckaert

The entropy problem no one talks about

Entropy is the foundation of every cryptographic operation in modern identity systems. Key generation, nonce derivation, signature randomness, session token creation: all of these depend on a source of unpredictability that an adversary cannot observe, reproduce, or influence.

The industry has spent decades refining entropy sources. Hardware random number generators (HRNGs) sample thermal noise from silicon. Quantum random number generators (QRNGs) measure photon arrival times or vacuum fluctuations. Operating systems blend interrupt timing, disk seek latency, and user input into kernel entropy pools.

These sources are effective. They are also, without exception, terrestrial. They exist within the physical and electromagnetic environment that an attacker occupies. A sufficiently motivated adversary, one with nation-state resources, sophisticated electromagnetic shielding, or physical proximity to the entropy source, can observe, model, or influence the conditions under which that entropy is generated.

This is not a theoretical concern. Side-channel attacks against hardware random number generators have been demonstrated in academic literature since the early 2000s. Electromagnetic emanation analysis can extract information about internal chip states. Power analysis can reveal correlations between entropy generation and observable electrical characteristics. Even QRNGs, which sample genuinely quantum phenomena, are vulnerable to implementation-level attacks such as detector blinding and Trojan-horse attacks against the optical path.

The common thread is proximity. Every terrestrial entropy source can be approached, probed, and potentially compromised because it exists in the same physical domain as the adversary.

Orbital physics

Low Earth Orbit (LEO) satellites operate at altitudes between approximately 200 and 2,000 kilometers above the Earth's surface. At these altitudes, their orbital velocities range from about 7.8 kilometers per second at 200 kilometers to about 6.9 kilometers per second at 2,000 kilometers. A single LEO satellite completes a full orbit in roughly 88 to 127 minutes, depending on its altitude.

This velocity creates a continuously shifting relationship between the satellite and any ground-based observer. The Doppler effect, the change in observed frequency of a signal emitted by a moving source, produces measurable frequency shifts that vary as a function of the satellite's position, velocity vector, and the observer's location on Earth. For a satellite transmitting at L-band frequencies (approximately 1.5 GHz), the Doppler shift can range from approximately +40 kHz to -40 kHz as the satellite passes from horizon to horizon.

The critical property of this Doppler shift, from an entropy perspective, is that it is non-deterministic at the precision required for cryptographic seeding. While the orbital trajectory of a satellite can be predicted with high accuracy using Two-Line Element (TLE) sets and SGP4/SDP4 propagation models, the actual observed Doppler shift at a specific receiver at a specific instant is influenced by factors that cannot be predicted with cryptographic precision:

• Ionospheric variability: The ionosphere introduces frequency-dependent delays and phase shifts that vary with solar activity, geomagnetic conditions, and local electron density. These variations are measurable but not predictable to the bit-level precision required to reconstruct an entropy seed.

• Tropospheric refraction: Temperature, humidity, and pressure gradients in the troposphere alter the signal path, introducing sub-microsecond timing variations that are locally unique.

• Multi-path interference: Reflections from terrain, buildings, and other structures create constructive and destructive interference patterns that are unique to the receiver's exact position and the satellite's instantaneous geometry.

• Constellation geometry: Modern LEO constellations (Starlink, OneWeb, Kuiper, Iridium NEXT) deploy thousands of satellites in multiple orbital planes. At any given moment, a ground receiver may observe signals from dozens of satellites simultaneously. The relative geometry of this visible constellation changes continuously, producing a time-varying spatial signature that is unique to the observer's location.

• Handover timing: As LEO satellites move across the sky, the ground terminal hands off from one satellite to another. The precise timing of this handover, measured at the microsecond or sub-microsecond level, depends on signal strength thresholds, constellation scheduling algorithms, and atmospheric conditions. The handover instant is unpublished, unpredictable to external observers, and irreproducible.

Taken individually, each of these factors introduces a small amount of uncertainty. Taken together, they produce a composite signal rooted in orbital mechanics, atmospheric physics, and electromagnetic propagation, which is hard to reproduce from a different location. How much usable entropy it holds would have to be measured; it has not been characterized for cryptographic use.

The architecture of orbital entropy

An identity system that incorporates LEO satellite signals as an entropy source does not replace existing entropy infrastructure. It layers orbital entropy on top of classical and quantum sources, creating a hierarchy of tiers with different physical origins.

Consider the following four-tier model:

Each tier is independently valuable. When the tiers are mixed correctly, an adversary must predict every tier to predict the output. The orbital tier is hard to observe remotely because it depends on conditions along a path from space to one receiver. The receiver itself is on the ground, though, and is as reachable as any other device.

Doppler shift as a cryptographic primitive

The Doppler shift from a LEO satellite is a deterministic physical measurement, not a random number. However, the measurement residual, the difference between the predicted Doppler shift (from published ephemeris data) and the actual observed Doppler shift (at the receiver), contains genuine unpredictability.

This residual is the entropy. It arises from the aggregate of all the atmospheric, geometric, and electromagnetic factors that are not captured in the orbital prediction model. At fine enough precision, these residuals cannot be predicted from publicly available data, are hard to reproduce without being near the measurement location, and change continuously as the satellite moves.

Extracting cryptographic entropy from Doppler residuals would follow the usual pattern for physical entropy sources:

1. Observe: Measure the Doppler shift from one or more LEO satellites at the receiver location.
2. Predict: Calculate the expected Doppler shift using published TLE data and an SGP4 propagator.
3. Subtract: Compute the residual (observed minus predicted).
4. Condition: Apply a randomness extractor (e.g., a cryptographic hash or a min-entropy conditioner conforming to NIST SP 800-90B) to the residual to produce uniformly distributed output bits.
5. Mix: Combine the conditioned orbital entropy with entropy from other tiers (QRNG, HRNG, carrier signals) using a domain-separated HMAC or XOF construction.

The output of this process is a seed that incorporates entropy from orbital mechanics, atmospheric physics, and the unique electromagnetic environment of the receiver's location. The seed is then used in standard cryptographic constructions: key derivation, nonce generation, or signature randomness.

Identity as an entropy consumer

Identity systems are entropy consumers. Every passkey registration generates a challenge nonce. Every verifiable credential issuance requires a unique identifier. Every hybrid PQC signature mixes a quantum-seeded nonce into the payload before signing. Every cross-reference binding produces a cryptographic hash that must be collision-resistant under adversarial conditions.

The security of these operations is only as strong as the entropy that seeds them. If an adversary can predict the nonce used in a passkey challenge, they can pre-compute the response. If an adversary can predict the randomness used in a signature, they can forge the signature. If an adversary can predict the seed used in key generation, they can derive the key.

Today's entropy sources are sufficient against today's adversaries. But the adversary model is changing. Cryptographically relevant quantum computers (CRQCs) threaten the mathematical foundations of classical cryptography. Nation-state actors invest in electromagnetic surveillance capabilities that can observe hardware entropy generation at a distance. Supply chain attacks can compromise the integrity of hardware random number generators before they are deployed.

Orbital entropy addresses the observation and supply-chain concerns in part, because much of it comes from conditions along the signal path rather than inside the receiver. It does nothing about CRQCs, which attack algorithms rather than randomness. A remote adversary cannot replicate the exact multi-path interference pattern at a receiver thousands of kilometers away, or predict the precise handover timing of a constellation they do not operate. An adversary near the receiver, or one who transmits spoofed satellite signals, is a different case.

The constellation advantage

A single LEO satellite provides a Doppler shift measurement that changes as the satellite transits the sky. The entropy contribution is significant but time-bounded: a single pass lasts 5 to 15 minutes, depending on the orbital altitude and the observer's latitude.

A constellation provides many more measurements. SpaceX's Starlink operates the largest LEO fleet, with thousands of active satellites. Eutelsat OneWeb operates a first-generation constellation of more than 600 satellites. Iridium NEXT operates 66 active satellites plus on-orbit spares. Amazon Leo (formerly Project Kuiper) has launched its first production satellites and has not yet started commercial service.

At any given moment, a ground receiver with a wide view of the sky can see many LEO satellites at once. Each satellite gives its own Doppler measurement along its own atmospheric path. The residuals are not fully independent, because satellites seen through the same ionosphere and the same local environment share some of the same errors, so the usable entropy grows with the number of satellites but by less than one full share per satellite.

As more constellations are deployed, more measurements are available at any ground location.

Handover timing as an entropy amplifier

LEO satellites move fast enough that a ground terminal must periodically switch from one satellite to another. In dense constellations, these handovers occur every few minutes. The exact timing of a handover depends on:

• The signal strength of the departing satellite (which depends on atmospheric conditions, antenna gain pattern, and satellite attitude)
• The signal strength of the arriving satellite (same dependencies)
• The constellation management algorithm's decision thresholds (which may incorporate load balancing, Quality of Service requirements, and predictive beam management)
• The propagation delay through the current atmospheric path

The handover instant, measured with sub-microsecond precision, is unpublished, externally unobservable (it occurs within the terminal's protocol stack), and irreproducible. It represents a timing measurement that depends on the convergence of orbital mechanics, atmospheric physics, and software state, three domains that are independently unpredictable at the required precision.

Handover timing can serve as an entropy amplifier: a source of additional unpredictability that is layered on top of Doppler residuals and constellation geometry to increase the total entropy rate of the orbital tier.

Non-terrestrial networks and 3GPP standards

Satellite links in cellular standards are not speculative. The 3GPP Non-Terrestrial Networks programme has been developing NTN specifications since Release 17, with significant enhancements in Release 18 and ongoing work in Release 19. NTN standards define how terrestrial cellular networks (5G NR) can operate over satellite links, including LEO, MEO, and GEO constellations.

The NTN architecture introduces several properties that are directly relevant to identity entropy:

• Timing Advance (TA) compensation: NTN requires sophisticated timing advance calculations to account for the propagation delay between a ground terminal and a moving satellite. These TA values are continuously updated and reflect the real-time geometric relationship between the terminal and the satellite.

• Doppler pre-compensation: The terminal must pre-compensate for the Doppler shift before transmitting to the satellite. This pre-compensation value is computed locally and depends on the terminal's position, the satellite's ephemeris, and the atmospheric conditions along the signal path.

• Ephemeris broadcast: NTN satellites broadcast their orbital parameters (position, velocity) to terminals, enabling local computation of geometric properties. The difference between the broadcast ephemeris and the actual observed signal properties contains the atmospheric and electromagnetic residuals that constitute orbital entropy.

These are standardized protocol mechanisms already specified for 5G NTN terminals. Whether a terminal exposes these values to applications is up to the device and its modem; if it does, the measurements come from the communication protocol itself rather than from additional spectrum or proprietary satellite access.

The adversary model

Every entropy source must be evaluated against a concrete adversary model. For orbital entropy, the relevant question is: what would an adversary need to do to predict or replicate the entropy output?

To predict the Doppler residual at a specific receiver at a specific instant, the adversary would need:

1. Precise knowledge of the ionospheric state along the signal path between the satellite and the receiver. This requires either a co-located ionospheric probe or a model of sufficient fidelity to predict electron density variations at sub-kilometer resolution and sub-second temporal resolution. No such model exists for arbitrary locations.

1. Precise knowledge of the tropospheric state along the signal path. Temperature, humidity, and pressure gradients at the required precision are not available from public meteorological data, which is typically sampled at hourly intervals and kilometer-scale spatial resolution.

1. Precise knowledge of the multi-path environment at the receiver location. This requires a complete electromagnetic model of the terrain, structures, and reflective surfaces within the Fresnel zone of the signal path. For a moving LEO satellite, the Fresnel zone sweeps across the landscape, making the multi-path environment time-varying and position-dependent.

1. Co-located measurement capability: Even with perfect models of the atmosphere and multi-path environment, the adversary would need a receiver at the exact location of the target receiver to measure the same signal. A receiver even meters away would observe different multi-path interference and different atmospheric path lengths.

This is the asymmetry of orbital entropy: a remote adversary cannot predict it at the required precision. Two limits remain. An adversary close to the receiver sees much of the same environment, and an adversary who transmits spoofed satellite signals, as is done against GNSS receivers, can try to control what the receiver measures. An orbital tier would need spoofing detection and would still be mixed with other sources.

PasskeyBridge's current entropy architecture

PasskeyBridge's current entropy architecture uses a four-tier QRNG-seeded hierarchy: Outshift (Cisco), QCi uQRNG (photonic quantum hardware), ANU QRNG (fallback), and local CSPRNG (emergency). This hierarchy seeds the nonces that PasskeyBridge mixes into its ML-DSA hybrid signatures. PasskeyBridge does not collect orbital or satellite entropy today.

An orbital tier is a research direction. If 5G NTN terminals expose the measurements to applications, an orbital tier could extend the existing hierarchy with a source that is hard to observe from a distance.

The integration path would reuse what exists: Doppler residuals, handover timing, and constellation geometry measurements would pass the same health tests that QRNG seeds pass today and be mixed with the existing tiers through a domain-separated HMAC construction. Mixed that way, a weak or manipulated orbital input cannot make the output more predictable than the other tiers alone.

Orbital entropy would add to QRNG rather than replace it, so that predicting the output requires predicting every tier.

Open questions

The identity industry has spent years debating algorithms: RSA versus elliptic curves, HMAC versus EdDSA, classical versus post-quantum. These debates matter. But they address the mathematical layer of cryptographic security while leaving the physical layer, the entropy that seeds every algorithm, under-examined.

Orbital entropy forces the conversation to a different plane. It asks not "which algorithm is strongest?" but "where does the randomness come from, and can an adversary reach it?"

For terrestrial entropy, the answer is yes: given sufficient resources, an adversary can approach, observe, or influence the source. For orbital entropy, much of the source lies along a path from space that a remote adversary cannot measure, but the receiver is still on the ground and its inputs can be spoofed.

The open questions are how much entropy the measurements hold once conditioned, whether terminals will expose them, and how well spoofing can be detected. Those need measurement before any orbital tier is used for keys or nonces.

Read the PQC Architecture Deep Dive → | Explore Spatially-Bound Identity → | View the Whitepaper →

Start free · Test the API