Fisher Information Limits of Satellite RF Fingerprint Identifiability for Authentication

Radio frequency fingerprinting can complement cryptographic satellite authentication, but an unknown channel can make hardware parameters unidentifiable. We derive a leading-order Fisher information matrix for joint in-phase/quadrature imbalance and memoryless power-amplifier parameters with a known or calibrated flat channel, followed by the complete matrix and nuisance projection for an unknown complex channel. The projection exposes a gain gauge: for ideal in-phase/quadrature balance and constant-modulus signals, the power-amplifier coefficient is absorbed into the channel; binary phase-shift keying has zero effective hardware rank, and symbol-rate quadrature phase-shift keying retains at most two real hardware coordinates. Full joint rank requires a nonlinear direction outside the nuisance spans of the channel and imbalance parameters, as can occur with known amplitude diversity. We state a local asymptotic distance relation based on Fisher information for pairwise parameter separation. Real-data features describe amplitude, phase dynamics, and moments of the in-phase and quadrature components rather than parameter estimates, so we evaluate them separately on two consecutive Iridium measurement campaigns. An enrollment-only, satellite-disjoint protocol gives every comparator the same accumulated inputs and reports satellite-level uncertainty. With ten-message accumulation, squared discrimination-ratio weighting in feature space obtains known/unseen-impostor areas under the receiver operating characteristic curve of 0.866/0.858, versus 0.710 with one message for unseen impostors. At an enrollment-calibrated 5% false-acceptance target, legitimate-satellite detection probability is 0.242 and unseen-impostor false acceptance is 0.041, exposing the gap between ranking and fixed-threshold detection.