Limitations

Active

Written by the author, before a reviewer writes it for them. A limitation stated plainly costs a paragraph; the same limitation found by a reviewer costs a revision cycle.

Scope

What the results hold for. A LEO broadband access network in which the operator can measure arrival time to roughly 10 ns and can observe each session from at least three satellites sharing a common transmit epoch. Under those two conditions, terminal position is recoverable to about 3 m and a single-beam adversary cannot shift its apparent position at all.

What they do not hold for. Everything else, and the two conditions are not minor.

Condition	If unmet
Simultaneous multi-satellite observation	Position and a free-running clock bias become jointly unidentifiable. The guarantee degrades to a clock-plausibility heuristic that a commodity TCXO defeats by tens of kilometres
Arrival time to roughly 10 ns	The resolution cell grows until legitimate neighbours are indistinguishable from a relay. At 100 ns, false positives reach 28.3% at 100 terminals per km²

Unmitigated threats

Threats named in the methodology that were not addressed, and what that costs.

Threat	Why not mitigated	Effect on conclusions
Real constellations use beam hopping, scheduling and proprietary handover logic the model omits	No access to operational scheduling behaviour	The evaluation is simulation on public ephemeris. It does not claim operational validation, and the numbers should be read as an upper bound on achievable geometry
No prior geometric method to compare against	The literature study that produced this project found none	The detector is characterised against its own operating envelope rather than a competitor. A reviewer may reasonably ask “compared to what?” and there is no strong answer
Timing precision at 10 ns is asserted as achievable, not demonstrated	Would require hardware the project does not have	The central operating condition rests on a value taken from link-budget reasoning rather than measurement
Estimator uses delay only	Doppler is modelled but not wired into the solve	$\delta$ is likely pessimistic. Since $\delta$ binds the false-positive envelope, the operating envelope reported may be narrower than achievable

Claims a reader should not draw

That this detects bots. It does not. Geometry establishes where, not who or what. Automation running on a legitimate user’s own terminal at its true location is invisible to the method, exactly as it is to device attestation.
That it prevents Sybil attacks. It bounds them. An adversary willing to deploy $\lceil M/q \rceil$ physical terminals at distinct sites obtains $M$ identities and is not detected. The contribution is that the cost becomes linear in hardware rather than in compute.
That spoofing is impossible. It is impossible for a single-beam adversary. An adversary radiating independently timed phase-coherent beams, one per satellite, can synthesise any position exactly. That class is not defended against; the barrier is capability cost, not geometry.
That the global identity cap is meaningful. It is not, and the paper deliberately does not present one. A 100 × 100 km footprint holds roughly 1.4 million resolution cells at $\delta = 47$ m; a ceiling of that magnitude is operationally vacuous. The meaningful statement is per terminal.
That the numbers transfer to another constellation. They were measured on Starlink elements. The method is not Starlink-specific, but the resolution and visibility figures are.
That 10 ns is a mild requirement. It is the binding practical constraint and it is stated as such in the abstract, not buried.

What would remove each limitation

Each limitation with a route out of it belongs in future work; each without one is a boundary of the method.

Limitation	Removable by	Cost
Delay-only estimation inflates $\delta$	Folding Doppler into the solve; it constrains the same unknowns independently	Low. Already modelled in the code, needs wiring and re-validation
No operational validation	Measurement campaign against a real terminal and gateway	High. Needs operator cooperation or instrumented hardware
10 ns asserted, not demonstrated	Bench measurement of achievable arrival-time precision on representative hardware	Medium
No comparison baseline	Implementing an RSSI-based Sybil detector from the WSN literature as a straw baseline	Medium. Would be a weak comparison but better than none
Multi-beam adversary undefended	Exploiting the fact that independently timed beams still radiate from a common aperture, so their relative phase behaviour may be anomalous	High, speculative. Currently stated as future work, not claimed
Fixed terminals only	Extending the estimator to a motion model	Medium. Changes the estimation problem rather than extending it
On-terminal automation invisible	Nothing within this method. Geometry cannot distinguish a human from a script at the same location	Boundary of the method, not a gap