NTN · 2026-04-18

Orbital Slot as Entropy Source: Harvesting Positional Uncertainty from Satellite Constellations

By J. W. Bouckaert

Every cryptographic key, nonce, and salt in a modern identity system traces back to a finite set of entropy sources. For most production stacks that set is small: an on-die hardware random number generator (Intel RDSEED, ARM TrustZone Entropy Source), a kernel CSPRNG mixing pool, and—in higher-assurance deployments—an external photonic quantum random number generator (QRNG) appliance.

These sources are well-understood, well-studied, and almost universally co-located with the systems they protect. That co-location is also their structural weakness. Every byte of randomness produced inside a data center is generated within physical and electromagnetic reach of any sufficiently determined adversary.

There is one entropy source that does not share this property: the satellites themselves. Their positions, rather than their downlink signals or their carrier-phase noise. The actual, observed-versus-predicted location of a Low Earth Orbit satellite at a specific instant in time is, at the precision required for cryptographic seeding, fundamentally non-deterministic. And the physics that drives that non-determinism—atmospheric drag, solar radiation pressure, geomagnetic storm-induced thermospheric density spikes—operates 400 to 1,200 kilometers above any electromagnetic shield an attacker can erect.

This article examines orbital slot uncertainty as a formal entropy source. We compare its information-theoretic properties to QRNG and CSPRNG outputs, walk through the NIST SP 800-90B validation pipeline as applied to orbital residual streams, and document the architectural pattern PasskeyBridge uses to harvest, condition, and consume orbital entropy without adding latency to the authentication path.

Orbital slot uncertainty, defined

The term "orbital slot" has two common meanings in aerospace literature. The first refers to geostationary slot allocations governed by the ITU—the 0.1-degree longitudinal positions where commercial GEO satellites are licensed to operate. That is not what we mean here.

The second meaning, and the one relevant to entropy harvesting, is the time-evolving position a satellite occupies along its predicted orbital track. For a LEO satellite, that track is computed from a Two-Line Element (TLE) set propagated forward in time using the Simplified General Perturbations 4 (SGP4) model—or its deep-space variant SDP4 for longer-period orbits. The TLE format and SGP4/SDP4 propagator are the standard public-domain mechanism for predicting satellite positions, maintained and distributed by the U.S. 18th Space Defense Squadron through Space-Track.org and mirrored by CelesTrak.

The critical fact, well-documented in the aerospace literature, is that SGP4 is a general perturbations model: it accounts for the dominant secular and long-period orbital perturbations (J2, J3, J4 zonal harmonics; lunar-solar gravitational effects; basic atmospheric drag via the BSTAR coefficient) but does not—and cannot—model the high-frequency stochastic forcing that actually drives short-term position uncertainty.

Independent accuracy assessments of SGP4 against precision orbit determination ground truth show position errors that grow approximately linearly with prediction horizon. For typical LEO altitudes and a TLE less than 24 hours old, along-track errors are commonly cited at 1–3 km, with cross-track errors of several hundred meters and radial errors of 100–200 meters. Studies such as "An Accuracy Analysis of the SGP4/SDP4 Model" (Wei & Zhao, 2009) catalog these residuals across thousands of objects.

Those residuals are the signal we want to harvest, not noise to eliminate.

Sources of uncertainty

The forcing functions that cause an actual satellite position to diverge from its SGP4 prediction are well-characterized physics, but the realized values at any given instant cannot be predicted to bit-level precision.

Atmospheric drag variance

Below approximately 600 km, atmospheric drag is the dominant non-conservative force acting on a LEO satellite. The drag force depends on local atmospheric density, which is driven by solar EUV flux (commonly indexed by F10.7 cm radio flux), geomagnetic activity (Kp/Ap indices), and—on multi-day timescales—the satellite's own ballistic coefficient and attitude.

The May 2024 Gannon geomagnetic storm provided a vivid empirical demonstration. A study from MIT (Parker & Linares, 2024) analyzing the storm's impact found that thermospheric density at LEO altitudes spiked dramatically over a 48-hour window, causing measurable orbital decay across the LEO catalog and forcing operators to issue collision-avoidance maneuvers across multiple constellations. Storm-induced density variations can reach factors of 2–10× nominal values, and the timing of the density spike at any specific orbital location is not predictable from publicly available models with the accuracy required to reconstruct the resulting position residuals.

Solar radiation pressure

Above approximately 800 km, solar radiation pressure (SRP) becomes a significant perturbation. SRP magnitude depends on the satellite's effective area-to-mass ratio, surface optical properties, and instantaneous geometry relative to the Sun. SRP also depends on solar activity—the same EUV variability that drives drag also modulates the photon flux. NASA's Goddard Space Flight Center maintains extensive technical references on radiation pressure modeling for spacecraft operations.

For high-LEO and MEO satellites, SRP introduces position uncertainty on the order of tens to hundreds of meters per day that cannot be removed by improving the propagation model alone, because the underlying solar flux is itself stochastic on the relevant timescales.

Third-body and tidal effects

Lunar and solar gravitational perturbations are deterministic in principle, but their realization in the truncated SGP4 force model introduces residual error that grows with prediction horizon. Earth tides—both solid-Earth tides and ocean tides—introduce additional sub-meter to meter-level position perturbations that are not captured in SGP4 at all.

The information-theoretic view

For an entropy source to be cryptographically useful, it must satisfy three properties:

1. Unpredictability. An adversary observing all prior outputs and possessing complete knowledge of the source's design must not be able to predict the next output with non-trivial probability.
2. Independence. Successive outputs must not exhibit statistically detectable correlation that would allow recovery of one output from observation of others.
3. Min-entropy lower bound. The worst-case unpredictability—measured as min-entropy—must meet a defined threshold under NIST SP 800-90B, the U.S. federal standard for entropy source validation.

Orbital residual streams—defined as the time series of (observed minus SGP4-predicted) position vectors—satisfy all three properties when sampled correctly. The unpredictability derives from the underlying atmospheric and solar physics. The independence derives from the chaotic nature of the upper-atmosphere coupling. The min-entropy is established empirically through the SP 800-90B validation pipeline.

Comparing orbital entropy to QRNG and CSPRNG

The table below summarizes the three entropy source classes against the dimensions that matter for production identity systems. All values reflect typical production deployments rather than theoretical maxima.

Orbital entropy is a layer on top of CSPRNG and QRNG, replacing neither, and it brings physical separation from the attacker as a defensive property no other source provides at commodity cost.

Min-entropy distribution across 30 days

The chart below shows the distribution of conservative min-entropy estimates produced by the SP 800-90B non-IID estimator suite (the smallest of the ten estimators, per the standard's guidance) across 30 days of orbital residual sampling on a single ground receiver in our pre-production validation environment. Bars indicate the share of one-hour windows that fell into each min-entropy band.

The distribution is right-skewed around 0.55 bits/byte raw, comfortably above the 0.5 threshold most reviewers consider the minimum acceptable for use as an entropy source after conditioning. The conditioning step—HMAC-SHA-256 with a domain-separated key—amplifies this to effectively full 8.0 bits/byte in the output stream, per the SP 800-90B vetted conditioning component analysis.

Sampling architecture

The end-to-end orbital entropy harvesting pipeline used by PasskeyBridge has five stages. Each stage is independently testable and individually attributable to a public standard or specification.

The entire pipeline runs as an asynchronous reseed loop on the entropy producer side. Authentication operations on the consumer side draw from the conditioned DRBG state and never block on the orbital sampler. This is the architectural detail that makes orbital entropy practical: it adds zero latency to the verification path.

Position in the entropy hierarchy

PasskeyBridge does not replace QRNG or CSPRNG sources with orbital entropy. We layer them. The four-tier hierarchy described in our earlier piece on LEO satellite signals as a ground-based entropy replacement gets a refinement here: orbital slot residuals occupy the fourth tier and are XOR-combined with the lower tiers in the conditioning step. A compromise of any single tier never collapses the output keystream.

The full hierarchy used in production:

XOR-combining tiers preserves min-entropy: the combined stream has min-entropy at least equal to the maximum of any contributing tier, and any non-correlated uplift from additional tiers is additive in the worst-case analysis. This is a direct application of the "leftover hash lemma" formalism that underpins SP 800-90B's recommendations on multi-source entropy combination.

For a deeper treatment of how this layered approach defends against harvest-now-decrypt-later attacks, see our companion analysis. The same hierarchy also feeds the QRNG-DRBG nonce generator referenced throughout the hybrid signature and post-quantum identity stack documentation.

Limits

Orbital entropy is a defensive layer. Three honest limitations are worth stating explicitly.

Throughput is bounded. A single ground GNSS receiver sampling at 1 Hz produces on the order of 1–10 KB of conditioned entropy per second. That is more than sufficient for nonce, salt, and key generation, but it is orders of magnitude below what a high-volume randomness consumer (e.g., a TLS termination edge) would need. Orbital entropy is for seeding and reseeding; bulk randomness still comes from the CSPRNG it seeds.

Receiver compromise remains in scope. If an attacker physically compromises the ground receiver, they observe the same residuals the legitimate consumer observes. The defense is the layered XOR with QRNG and HRNG: compromising one tier never defeats the composite. This is also why the receiver and the entropy consumer should ideally be on different hosts, with the conditioned output transported over an authenticated channel.

Catalog manipulation is the residual concern. A nation-state adversary could in principle inject false TLE data into the published catalogs. PasskeyBridge mitigates this by cross-validating TLE data from multiple independent sources (Space-Track plus CelesTrak plus, where available, ESA's DISCOSweb) and by sampling from a wide constellation rather than relying on any single satellite.

Identity's threat model

Most discussion of entropy sources happens in the context of TLS, key wrapping, or cryptocurrency wallets. Identity has a different threat model.

In an identity system, the entropy consumer is a continuously running stream of nonce derivations, session token mints, challenge generations, and DRBG reseeds, far from a one-shot key-generation event. A weakness in the entropy source is therefore a persistent, low-grade information leak that compounds over millions of authentications rather than a single-event failure.

The blast radius of a compromised entropy source in an identity system includes:

• Predictable WebAuthn challenge values that enable replay against captured passkey signatures
• Predictable session token IDs that enable session enumeration
• Predictable DPoP nonces that defeat the DPoP token-binding model entirely
• Predictable Verifiable Credential nonces that enable presentation replay

Every one of these failure modes is invisible to the user, invisible to the relying party, and invisible to most monitoring stacks. The only defense is a defense-in-depth entropy architecture where no single source compromise propagates into the consumer.

Orbital slot entropy is the layer that makes the architecture defensible against an adversary who has reached the data center.

Standards alignment

For organizations subject to FIPS 140-3 validation or SOC 2 Type II audit, the orbital entropy pipeline is structured to align with established requirements:

• NIST SP 800-90A Rev. 1—DRBG mechanism (Hash_DRBG with SHA-256)
• NIST SP 800-90B—Entropy source validation (continuous health tests, IID/non-IID estimator suite, restart tests)
• NIST SP 800-90C—DRBG construction (chaining of vetted conditioning function and Hash_DRBG)
• FIPS 140-3 Section 7.8—Random bit generation requirements
• ISO/IEC 18031:2011—Random bit generation (international harmonization)

Validation evidence is captured in the audit trail described in our zero-knowledge audit trail piece—every entropy reseed event records the source tier, conservative min-entropy estimate, and conditioning parameters without exposing the underlying entropy bytes.

For teams already deploying post-quantum signatures or evaluating physics-anchored identity tokens, orbital entropy is the natural complement: physical-world anchoring on the entropy side, lattice-based cryptography on the signature side. Both move the system away from the assumption that the data center is a trusted environment.

Position

The default identity stack treats entropy as a solved problem. It is not. Every byte of randomness produced inside a server rack is generated within physical reach of any adversary who has reached that rack. Side-channel attacks against on-die HRNGs are documented. Detector-level attacks against QRNG appliances are documented. The assumption that "entropy comes from the chip and the chip is trusted" is the assumption that needs to change.

LEO satellite positional uncertainty offers a structurally different defensive property. The physics that drives the unpredictability—atmospheric drag variance, solar radiation pressure, geomagnetic storm-induced thermospheric coupling—operates above the electromagnetic envelope that any terrestrial adversary occupies. The entropy is real, the standards alignment is documented, and the throughput is more than sufficient for identity-system seeding and reseeding.

It does not replace the entropy sources you are already using. It joins them as a fourth tier whose compromise vector does not overlap with any of the three below it. That non-overlap is what makes it worth adding. In a defense-in-depth entropy architecture, the value of an additional source is exactly equal to the independence of its failure mode from the sources you already have.

For identity systems—where every nonce, every challenge, and every session token traces back to the same upstream entropy pool—that independence is worth reaching for.

If you want to see how orbital entropy integrates into a production identity stack, the PasskeyBridge SDK documents the consumer-side API. The whitepaper covers the layered hierarchy in more depth, and the solutions page maps the entropy architecture to specific compliance frameworks.

Start free · Test the API