From Spectrum Regulation to Computational Enforcement: An Auditable Governance Architecture for Adaptive Spectrum Sharing

1. Introduction

• The governance problem is translating regulatory constraints into machine-readable, executable, traceable, and revisable decisions without losing institutional meaning.The paper asks how to enforce international and national constraints while retaining an auditable chain between policy, interference assessment, authorization, and enforcement.Source: §1 · pp. 3–4

sessment, authorization, and enforcement?

• SPECTRA-GOV connects treaty-level coordination, national adaptive licensing and incumbent protection, and real-time enforcement through TLAGA.The paper presents an auditable governance-to-execution workflow, with Layer 2 as its principal quantitative focus.Source: sessment, authorization, and… · p. 4
• Per-operator provenance distinguishes an operator contributing to an aggregate violation from one selected for suspension.The V2 evaluation adds controlled contention, paired baseline counterfactuals, and policy perturbations to the preserved original implementation.Source: sessment, authorization, and… · pp. 4–5
• The study limits its claims: propagation is simplified, thresholds are calibration fixtures, selective authorization is greedy, and latency is measured in-process.Layers 1 and 3 are evaluated architecturally or mechanistically, not as operational systems.Source: sessment, authorization, and… · pp. 4–5

2. Background and Related Work

• The paper organizes related work around dynamic sharing, institutional coordination, and trust and accountability rather than treating these debates as interchangeable.Its contribution is framed as an integrated architecture and policy-to-decision provenance model.Source: §2 · p. 6, §2.4 · p. 8

2.1. Dynamic spectrum sharing and spectrum governance

• Dynamic authorization is an established technical precedent; SPECTRA-GOV focuses on connecting it to regulatory authority, cross-jurisdictional incumbent records, geographic exclusion, provenance, and enforcement.The paper explicitly says dynamic authorization itself is not new.Source: §2.1 · p. 6

2.2. Institutional and regulatory coordination

• SPECTRA-GOV addresses the computational interface between institutions and automated decisions, making relevant rules explicit, versioned, executable, and auditable.It does not claim to solve the underlying legal or political coordination problem.Source: §2.2 · p. 7

2.3. Trust, accountability, and auditable digital governance

• The architecture links computational decisions to their inputs and rules, while the experiments evaluate reproducibility and provenance rather than community accountability or institutional legitimacy.The paper distinguishes reproducibility from accountability to affected communities and procedural fairness.Source: §2.3 · pp. 7–8

2.4. Position of the present contribution

• The novelty claim is at the architecture and governance-execution level, not in a propagation formula, blockchain primitive, or selective-admission heuristic.The selective SPECTRA-GOV mechanism matches the Dynamic-SAS baseline in access and protection, so selective admission is not claimed as an exclusive algorithmic contribution.Source: §2.4 · p. 8

3.1. Governance-to-execution flow

• The governance-to-execution flow applies international rules through national policy and licensing constraints to transmissions, then records authorization and enforcement decisions for audit and review.The figure is described as a conceptual flow from institutional rules to machine-executable constraints and provenance.Source: §3.1 · p. 9

3.2.1. Layer 1: ITU treaty harmonization

• Layer 1 represents international coordination, while the versioned incumbent registry supplies a machine-readable representation for downstream authorization.The registry is a computational abstraction, not a substitute for ITU Radio Regulations or bilateral coordination.Source: §3.2.1 · p. 10

3.2.2. Layer 2: national adaptive licensing

• Layer 2 is the principal quantitative focus: its licensing logic evaluates aggregate interference against incumbent constraints and supports geographic exclusion and time-bounded authorization.PIEZ supplies interference-margin and geographic-exclusion mechanisms; SATs represent authorizations that enforcement can suspend or revoke.Source: §3.2.2 · p. 11
• When aggregate interference exceeds the effective threshold, selective authorization suspends the largest contributors until the remaining aggregate complies.This is a greedy mechanism, not an optimization theorem, and it does not establish fairness, economic optimality, or strategy-proofness.Source: §3.2.2 · pp. 11–12

3.2.3. Layer 3: real-time enforcement

• Layer 3 includes the enforcement engine and regulatory sandbox, but the paper measures only an in-process detection-and-decision path under a controlled token registry.It excludes live traffic, network transport, cryptographic dispatch, and regulator infrastructure.Source: §3.2.3 · p. 12

4.1. Evaluation principles; 4.2. Reference incumbent fixture

• The evaluation separates legacy regression, mechanism validation, and controlled V2 experiments; its five-incumbent fixture is a fixed test environment, not an empirical tolerance estimate.This separation is intended to prevent V2 stress-test results from replacing the V1 paper result.Source: §4.1 · pp. 12–13, §4.2 · p. 13

4.3. Propagation and distance model

• V2 replaces the mixed latitude-longitude-altitude Euclidean norm with a geodesic horizontal distance combined with vertical separation to calculate slant range.The propagation implementation is ITU-R-informed, not a complete regulatory-grade P.452 or P.619 implementation.Source: §4.3 · p. 14

4.4. V1 legacy Monte Carlo regression

• The regression reproduces the original aggregate headline: a 100.00% incumbent protection rate, not a real-world protection probability, because the regional frequency configuration does not deliberately generate overlap with protected incumbent bands.Source: §4.4 · pp. 14–15

4.5. V2 contention scenario taxonomy

• Seven controlled V2 contention classes use 10,000 scenarios each; the adversarial class adds a full-overlap aggressor within approximately 2 km of the target incumbent.That class is a mechanism-level stress condition, not a claim about real-world attack frequency or geography.Source: §4.5 · pp. 15, 17

4.6. Mechanism validation

• The corrected co-located-jammer validation produces a violation against a 10^-3 mW threshold, while the earlier 0.001-degree-offset example remains compliant under the corrected distance model.These constructed cases validate the decision path, not propagation against field measurements.Source: Table 3 · p. 17, §4.6 · p. 17

4.7. Baseline comparison

• The baseline experiment compares five authorization semantics, contrasting population-wide denial on aggregate violation with selective suspension of contributing operators.It defines access opportunity and incumbent protection as distinct outcome measures and includes a static-license counterfactual with entrants added to the original population.Source: §4.7 · pp. 17–18

4.8. Policy frontier

• The policy-frontier experiment varies the incumbent threshold multiplier while reusing physical interference calculations, isolating policy sensitivity from new scenario realizations.The design prevents a threshold change from being conflated with a new random realization of physical interference.Source: §4.8 · p. 18

4.9. Regulatory-change and provenance experiments; 4.10. Enforcement-latency benchmark

• Regulatory-change and provenance experiments use separate 600-scenario populations; the audit chain records operator contributions, thresholds, compliance, suspension, and marginal contribution.The latency benchmark includes only detected violations and tests a constructed aggressor token, isolating an in-process suspension path rather than a production registry or network stack.Source: §4.9 · p. 19, §4.10 · p. 19

5.1. V1 regression and the evidentiary baseline

• The V1 regression reproduced 100.00% incumbent protection, a 4.4× spectrum-efficiency quantity, a 47 ms enforcement design target, and a $4.0T illustrative value.This confirms preservation of the original computational state, not protection under arbitrary contention; the original configuration did not deliberately generate protected-band overlap.Source: §5.1 · p. 20

5.2. Corrected mechanism validation

• The corrected geodesic model detected a protection-threshold violation for an explicitly co-located jammer, while LEO-altitude and original-offset cases remained below the threshold.This shows detection when sufficient physical proximity is constructed, correcting the interpretation of the earlier V1 offset example as a violation case.Source: §5.2 · p. 21

5.3. V2 contention stress test

• In the V2 stress test, adversarial close-proximity S4 produced the largest protection reduction, while spectral overlap alone caused relatively small reductions under reference parameters.The stress test covered 10,000 scenarios per class; the bootstrap intervals quantify sampling uncertainty within controlled scenarios, not real-world protection probability.Source: §5.3 · pp. 22–23

5.4. Baseline comparison

• Selective authorization restored 100% incumbent protection at reduced access opportunity, while population-wide aggregate gating offered higher access opportunity but allowed a small protection loss.Source: §5.4 · p. 25

5.5. Policy frontier

• At the 0.1 multiplier, selective authorization retained 96.94% protection and 62.41% access opportunity, while population-wide authorization maintained 100% protection and approximately 90.95% access.These results are conditional on the reference incumbent fixture, population generator, and greedy selection policy, not a universal guarantee.Source: §5.5 · p. 25

5.6. Regulatory change and provenance

• The current RCIA experiment reports a 5.99% operator reconfiguration rate at 0.1× and the same rate at tighter multipliers, while a separate S7 moderate-overlap policy-change experiment reports 31.38% at 0.1×.These values are not directly interchangeable because the experiments use different scenario populations and answer different questions.Source: §5.6 · pp. 25, 27

5.7. Enforcement-latency benchmark

• In-process enforcement latency increased approximately 60-fold from one to 500 operators, consistent with candidate-population scaling in detection while the decision stage remained comparatively small.These are software execution measurements conditional on detected violations and cannot be generalized to an operational platform without additional system-level measurements.Source: §5.7 · p. 28

5.8. Policy conformance and mutation testing

• All seven tested policy mutations conformed to their predefined directional expectations, yielding a 100% policy-conformance rate and zero surviving mutations.The test is a narrow regression of the decision path and does not cover QoS-tier effects, jurisdictional routing, or temporal validity.Source: §5.8 · pp. 28, 30
• Conflicting-rule interactions were outside the scope of this experiment.Source: §5.8 · p. 30

• The evidence is organized into four levels.Source: §6 · p. 30
• The distinction among evidence levels is more than a reporting convention.Source: §6 · p. 30
• Evidence levels determine what can legitimately transfer from the computational experiment to policy.Source: §6 · p. 30

7.1. Machine-readable governance as an institutional interface

• Machine-readable rules can specify protected frequencies, interference thresholds, geographic exclusions, authority, and validity, enabling consistent execution and rule-version evidence.Source: §7.1 · p. 30
• Adaptive authorization can recompute eligibility when the protected environment changes, and the policy frontier demonstrates computational parameter sensitivity without establishing a socially appropriate threshold.Source: §7.1 · p. 30

7.2. Auditability requires more than reproducibility

• V2 provenance records operator-level grounds for authorization decisions, distinguishing a suspended constructed aggressor from an authorized harmless background operator.Source: §7.2 · p. 33

7.3. Selective authorization exposes the protection–access trade-off

• Selective authorization achieves 100% protection and 89.43% access opportunity in S3, while aggregate dynamic gating achieves 99.91% protection and 99.53% access.Source: §7.3 · p. 33
• Selective Dynamic-SAS and SPECTRA-GOV results are identical, so selective admission is not a unique SPECTRA-GOV algorithmic contribution; the contribution is its governance and provenance setting.Source: §7.3 · p. 34

7.4. Cross-jurisdiction coordination remains institutional

• A shared computational representation does not create legal authority or resolve rule conflicts, cross-border disputes, record control, or community participation.Source: §7.4 · p. 34

7.5. Regulatory experimentation

• A regulatory sandbox could compare policy semantics on a fixed physical population, but a synthetic-population frontier cannot by itself establish social welfare or fairness.Source: §7.5 · pp. 34–35

7.6. What the V2 audit changes about the research claim

• The V2 audit corrected issues weakening the evidentiary chain, making the computational evidence more defensible without proving the architecture.Source: §7.6 · p. 35

8.1. Propagation model

• The propagation model is simplified and is not a complete regulatory-grade implementation of ITU-R P.452 or P.619; correcting geodesic distance does not empirically validate it.Source: §8.1 · p. 35

8.2. Reference incumbent thresholds

• Experiments establish behavior relative to configured reference thresholds, not protection guarantees for real radar, railway, AIS, or aviation incumbents.Source: §8.2 · p. 36

8.3. Scenario construction

• The deliberately constructed overlap and proximity scenarios remain synthetic; S4 is intentionally severe, and S6 varies populations across draws rather than modeling within-scenario joining and leaving.Source: §8.3 · p. 36

8.4. Selective authorization

• The greedy largest-contributor-first selective rule is not proven optimal, fair, incentive compatible, or strategy-proof, and alternative policy criteria could change the trade-off.Source: §8.4 · p. 36

8.5. Governance completeness

• The conformance experiment does not fully exercise QoS tiers, jurisdictional interactions, temporal validity, or conflicting rules across authorities, all needed for a full regulatory implementation.Source: §8.5 · p. 36

8.6. Enforcement latency

• Latency uses in-process Python timing conditional on detected violations and a single SAT token, excluding network, cryptographic, database, hardware, infrastructure, and human delays.Source: §8.6 · p. 37

8.7. Economic quantities

• The 4.4× spectrum-efficiency quantity and $4.0T economic value remain historical illustrative or calibrated quantities, not independently validated by V2.Source: §8.7 · p. 37

8.8. Accountability and affected communities

• Computational provenance identifies the rule and data behind a decision but does not establish contestability, equitable outcomes, or legitimate institutional authority.Source: §8.8 · p. 37

9. Reproducibility and Research Integrity

• V2.3 passed 196 automated tests, Python compilation, and a whitespace check; rerunning the V1 regression reproduced the original as-shipped aggregate outputs.Source: §9 · pp. 37–38
• The V2 evidence snapshot records 10,000 scenarios per S0–S6 contention class, plus 3,000 scenarios per class and 2,000 bootstrap resamples for uncertainty analysis.Experiment-specific counts include 2,000 paired scenarios for the primary baseline comparison, 1,000 per class for the all-class baseline comparison, 600 for RCIA and S7 policy-change experiments, and 2,000 for the policy frontier.Source: §9 · p. 38
• Machine-readable outputs are the source of record for V2 values, with RESULTS_V2.md retained as historical provenance and RESULTS_V2_CURRENT.md as the current evidence snapshot.Source: §9 · p. 38
• No V1 result was silently rewritten to favor V2; corrections are incremental and versioned, and the original paper implementation remains separately identifiable.Source: §9 · p. 38

10. Conclusion

• V2 extends rather than replaces the original implementation, with corrected geodesic distance, explicit interval overlap, deterministic seeds, safety margins, and machine-readable provenance.The evaluation also includes handling for geographic exclusion, registry updates, and consistent simulated time.Source: §10 · p. 39
• The implementation operated under controlled contention and distinguished a deliberately constructed violation from a compliant case.Source: §10 · p. 39
• Protection remained near 100% in most controlled S0–S6 classes but fell to 90.00% in the deliberately adversarial close-proximity class.The experiments report conditional findings across controlled scenarios.Source: §10 · p. 39
• Selective authorization achieved 100% incumbent protection and 89.43% access opportunity, versus 99.91% protection and 99.53% access under population-wide dynamic gating.The selective Dynamic-SAS and SPECTRA-GOV variants produced identical results.Source: §10 · p. 39
• Identical selective-variant results bound the novelty claim: selective admission is not presented as uniquely novel, while the contribution integrates policy execution, provenance, adaptive authorization, and cross-layer governance.Source: §10 · p. 39
• The study does not establish real-world protection probability, regulatory-grade propagation accuracy, fairness, strategic robustness, field enforcement latency, economic benefit, or institutional adoption.Future work includes validated propagation calculations, empirical geographic and traffic distributions, temporal dynamics, strategic behavior, jurisdictional interactions, and stakeholder validation.Source: §10 · pp. 39–40
• SPECTRA-GOV is a computational governance reference architecture and auditable research platform that makes rule-to-decision reasoning explicit enough to test, correct, reproduce, and challenge.Source: §10 · p. 40

Declaration of competing interest

• The authors declare no known competing financial interests or personal relationships that could have influenced the reported work.Source: Declaration of competing interest · p. 40

Data and code availability

• The project repository provides the reference implementation, tests, audit procedures, machine-readable V2 results, and reproducibility documentation, with versioned Git history.The original implementation is tagged v0.1.0-paper, and the post-audit checkpoint is tagged v0.2.3-post-audit-hardening.Source: Data and code availability · pp. 40–41