Benchmarks · lanes H, G, E
Quantum vs classical
Seven classical payment constructions against ideal S-money, compared on what each one assumes before any timing is read. There is no single score: every pair is either qualified by named concessions or incomparable.
- Classical records
- 23MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- Prevent double spend
- 15
- Qualified pairs
- 7
- Incomparable pairs
- 4
- Comparable pairs
- 0none: assumptions never fully match
Constructions
Double spending, latency and what each design assumes
Each classical row is one lane H BenchmarkRecord. Expand a row for its concessions, record id and the command that regenerates it. Sorting never merges columns into a ranking.
12 of 24 rows · classical latencies are MODELLED lower bounds (light-cone propagation + declared processing) on SYNTHETIC networks
| Trust · hardware | Privacy | ||||||
|---|---|---|---|---|---|---|---|
| Bounded (theorem) | not on matched topology | not on matched topology | Bob honest for eps_unf; privacy defined against Bob guessing b before presentation | Late choice, any site | eps_priv-private: Bob's guess of b before presentation <= 1/2^M + eps_priv | Reference | |
| Succeeds | 10 µs | — | none: copyable bearer string, offline verifiers, no shared state | Late choice, any site | none | IncomparableX11-H0 | |
| Prevented | 137 µs | — | honest online coordinator; authenticated merchant->coordinator channel (HMAC-SHA256) | Any site, online | none (coordinator learns token, merchant and site at spend time) | QualifiedX11-H1 | |
| Prevented | 10 µs | 83.4 µs | honest issuer relays commit; authenticated channels | Late choice, any site | none (issuer learns site at Stage-II commit) | QualifiedX11-H3-late-disclosed | |
| Prevented | 10 µs | 83.4 µs | honest-but-curious issuer; IDEAL 1-of-M OT functionality (real instantiation computational, NOT_IMPLEMENTED) | Late choice, any site | site hidden from issuer under ideal-OT receiver privacy (model) | QualifiedX11-H3-late-ot | |
| Prevented | 10 µs | — | honest issuer and site verifiers; authenticated provisioning | Site fixed at issuance | none (issuer learns site at issuance) | QualifiedX11-H3-preassigned | |
| Prevented | 137 µs | — | ML-DSA EUF-CMA (computational) + honest online spent ledger | Any site, online | none (coordinator learns token, merchant and site) | QualifiedX11-H4+H1 | |
| Prevented | 10 µs | — | EUF-CMA of ML-DSA (computational, post-quantum); offline verifiers hold issuer pk; site binding signed at issuance | Site fixed at issuance | none (issuer learns site at issuance) | QualifiedX11-H4+H3 | |
| Succeeds | 10 µs | — | EUF-CMA of ML-DSA (computational, post-quantum); offline verifiers hold issuer pk | Late choice, any site | none (serial linkable) | IncomparableX11-H4 | |
| Prevented | 20 µs | — | H5-TR: tamper-resistant secure element (no state extraction, cloning or counter rollback) | Late choice, any site | none modelled (device id + counter visible at clearing) | IncomparableX11-H5 | |
| Out of model | 20 µs | — | H5-TR: tamper-resistant secure element (no state extraction, cloning or counter rollback) -- VIOLATED in this run | Late choice, any site | none modelled (device id + counter visible at clearing) | No X11 pair | |
| Detected, not prevented | 20 µs | — | RSA (toy 512-bit) blind signature + cut-and-choose; bank honest for crediting; detection-only double-spend deterrence (identity revealed ex post) | Late choice, any site | payer untraceable if coin spent once (heuristic here: SHA-256 g is not bijective) | IncomparableX11-H6 |
- local validation (MODELLED)
- Stage-II commit before presentation
Log scale. Lower bounds: light-cone link delay plus declared processing (site 10 µs, coordinator 50 µs). S1 has no point here: its matched-topology timing is lane GH (not merged), so the two families never share this axis.
| Row | Quantity | ns |
|---|---|---|
| H0 negative control | local validation | 10000 |
| H1 coordinator | local validation | 136712.8190396304 |
| H3 late commit disclosed | local validation | 10000 |
| H3 late commit disclosed · commit | commit before presentation | 83356.4095197916 |
| H3 late commit ot private | local validation | 10000 |
| H3 late commit ot private · commit | commit before presentation | 83356.4095197916 |
| H3 preassigned | local validation | 10000 |
| H4 mldsa h1 ledger | local validation | 136712.8190396304 |
| H4 mldsa h3 site binding | local validation | 10000 |
| H4 mldsa none | local validation | 10000 |
| H5 assumption holds | local validation | 20000 |
| H5 assumption violated clone | local validation | 20000 |
| H6 cfn88 sec1 k16 | local validation | 20000 |
Matched topology
Quantum and classical on the same network
Not available yet
Matched-topology quantum runs (S-money on lane H's synthetic networks) belong to lane GH, which is in flight and not merged. Until then quantum and classical latencies are never placed on one axis.
expected source: lane GH: S-money runs on net-h-synthetic-2site-10km / 3site-100km
X11 comparability matrix
Concessions, column by column
Lane H's matrix compares each classical construction with the ideal S1 model on sixteen dimensions. QUALIFIED means comparable only after the listed concessions are priced in.
classical_h0_bearer[negative_control]
vss_money_kent_2022_ideal_2siteIncomparable — fails the double-spend goal or uses a different trust model| Dimension | Classical construction | S1 ideal (Kent 2022) |
|---|---|---|
| adversary power | any copier (no assumption needed to double spend) | unbounded dishonest Alice (information-theoretic), ideal lossless/noiseless model |
| quantum capability | none | Bob prepares BB84 qubits, Alice measures on receipt (no quantum memory) |
| issuer trust | irrelevant: double spend succeeds anyway | Bob honest for eps_unf; privacy defined against Bob guessing b before presentation |
| agent trust | honest site verifiers | trusted agents B,B0,B1 and A,A0,A1 controlling secure laboratories |
| trusted hardware | none | secure laboratories of trusted agents; no tamper-resistant user device |
| pre shared key bits | 0 | secure authenticated same-party channels from previously distributed keys (budget not given in Lemma 1) |
| preassignment of site | none (late choice) | none: b chosen in Stage II, but c=b XOR z must reach B_i in causal past of Q_i (Table I steps 7-9) |
| spender privacy | none (serial linkable) | eps_priv-private: Bob's guess of b before presentation <= 1/2^M + eps_priv |
| site count | 2 | 2 (2^M via Appendix H, not used here) |
| spacelike separation | both sites accept spacelike-separated copies | Q0,Q1 may be spacelike separated; validation without cross-checking |
| finality semantics | local accept; no conflict resolution | local validation at Q_b; settlement/irrevocability not specified |
| availability assumptions | offline; always available | Stage-II bit c relayed to both sites before presentation; ideal: no loss |
| proof certainty | negative control: double spend observed in model (1 extra accept) | PUBLISHED_THEOREM Lemma 1 eq.(1) (ideal model only) |
| security target | none met (C1 double spend succeeds) | eps_unf=(1/2+1/(2*sqrt2))^N |
| local verification metric | MODELLED local processing only | near-instant local check at Q_b; no paper latency number used |
| issuance included | serial list push to every site | Stage I quantum exchange + t,u distribution, arbitrarily early |
Latency case studies · lane G
Published topologies and a measured network
The Jiang et al. 2024 field topologies reproduce the paper's mean transaction times from fibre length alone. RIPE Atlas gives measured one-way delays (RTT/2) between real anchors.
- modelled (QMint)
- published
- light-cone lower bound
- measured (RIPE Atlas)
Error bars: published ± (Jiang), p05–p95 of RTT/2 (RIPE), min–max over 20 seeds (modelled on RIPE samples). Each case is its own topology; rows are not a matched quantum-vs-classical benchmark.
| Case · quantity | ns | Evidence |
|---|---|---|
| Jinan intra-city, 2.766 km fibre · S-money transaction time (modelled) | 15337 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| Jinan intra-city, 2.766 km fibre · Published mean transaction time | 15340 | PUBLISHED-AGGREGATE: Number quoted from a paper table or text |
| Jinan intra-city, 2.766 km fibre · Published classical cross-check (2 L / c_fibre) | 27660 | PUBLISHED-AGGREGATE: Number quoted from a paper table or text |
| Jinan intra-city, 2.766 km fibre · Light-cone lower bound (straight line / c) | 1420.9830455441274 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| Yiyuan–Mazhan inter-city, 60.54 km fibre · S-money transaction time (modelled) | 304207 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| Yiyuan–Mazhan inter-city, 60.54 km fibre · Published mean transaction time | 304200 | PUBLISHED-AGGREGATE: Number quoted from a paper table or text |
| Yiyuan–Mazhan inter-city, 60.54 km fibre · Published classical free-space reference (2 D / c) | 344000 | PUBLISHED-AGGREGATE: Number quoted from a paper table or text |
| Yiyuan–Mazhan inter-city, 60.54 km fibre · Light-cone lower bound (straight line / c) | 172119.07312224642 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| RIPE Atlas Helsinki → Tampere · One-way delay p50 (RTT/2) | 1947596 | MEASURED-LATENCY: Latency from real measurements |
| RIPE Atlas Helsinki → Tampere · S-money transaction time, median of 20 seeds | 1979622 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| RIPE Atlas Helsinki → Tampere · Light-cone lower bound (chord / c) | 517413.8543335968 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| RIPE Atlas Frankfurt → Tampere · One-way delay p50 (RTT/2) | 20381004.5 | MEASURED-LATENCY: Latency from real measurements |
| RIPE Atlas Frankfurt → Tampere · S-money transaction time, median of 20 seeds | 20401727.5 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
| RIPE Atlas Frankfurt → Tampere · Light-cone lower bound (chord / c) | 5211291.050926245 | MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement |
Paper topology
Jinan intra-city, 2.766 km fibre
- S-money transaction time
- 15.337µsMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- Published mean transaction time
- 15.34µs[15.33, 15.35]PUBLISHED-AGGREGATE: Number quoted from a paper table or text
- Published classical cross-check
- 27.66µsPUBLISHED-AGGREGATE: Number quoted from a paper table or text
- Light-cone lower bound
- 1.421µsMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- — timing/causality model only; says nothing about security
- — dT_tran = dT_proc + L_fibre / c_fibre (Jiang et al. 2024, Methods eq. (10)-(11))
- — secret validation record pre-distributed to the remote bank agent
- — fibre 2766 m at 2e+08 m/s
$PY -m pytest tests/unit/networks/test_networks_s_money.py -q
Jiang et al. 2024, https://arxiv.org/abs/2408.13063 (main text p.6 ('426 m', '2.766 km'); Fig.2(a) caption ('around 425 m'); Methods eq.(10)-(13); SI Table S2)
Paper topology
Yiyuan–Mazhan inter-city, 60.54 km fibre
- S-money transaction time
- 304.21µsMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- Published mean transaction time
- 304.2µs[304.19, 304.21]PUBLISHED-AGGREGATE: Number quoted from a paper table or text
- Published classical free-space reference
- 344µsPUBLISHED-AGGREGATE: Number quoted from a paper table or text
- Light-cone lower bound
- 172.12µsMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- — timing/causality model only; says nothing about security
- — dT_tran = dT_proc + L_fibre / c_fibre (Jiang et al. 2024, Methods eq. (10)-(11))
- — secret validation record pre-distributed to the remote bank agent
- — fibre 60540 m at 2e+08 m/s
$PY -m pytest tests/unit/networks/test_networks_s_money.py -q
Jiang et al. 2024, https://arxiv.org/abs/2408.13063 (main text p.6; Fig.2(b) caption; Methods eq.(14)-(15) p.15; SI Table S3)
Measured network
RIPE Atlas Helsinki → Tampere
- One-way delay p50
- 1.9476ms[1.7605, 2.2101]MEASURED-LATENCY: Latency from real measurements
- S-money transaction time, median of 20 seeds
- 1.9796ms[1.9158, 2.3863]MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- Light-cone lower bound
- 0.51741msMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- — timing/causality model only; says nothing about security
- — dT_tran = dT_proc + L_fibre / c_fibre (Jiang et al. 2024, Methods eq. (10)-(11))
- — secret validation record pre-distributed to the remote bank agent
- — one_way_ns = RTT/2: assumes symmetric forward/return paths; RTT includes end-host processing and queueing; NOT a physical lower bound and NOT a one-way measurement
- — clock error bound 1000 ns assumed (NTP-class)
$PY -m pytest tests/unit/networks/test_networks_ripe.py -q
RIPE Atlas msm 7714918 (RIPE Atlas REST API v2 (public measurement results)), retrieved 2026-10-11T21:15:00Z
Measured network
RIPE Atlas Frankfurt → Tampere
- One-way delay p50
- 20.381ms[20.188, 20.539]MEASURED-LATENCY: Latency from real measurements
- S-money transaction time, median of 20 seeds
- 20.402ms[20.283, 20.644]MODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- Light-cone lower bound
- 5.2113msMODELLED-LATENCY: Latency from a model (light-cone, queue model); not a measurement
- — timing/causality model only; says nothing about security
- — dT_tran = dT_proc + L_fibre / c_fibre (Jiang et al. 2024, Methods eq. (10)-(11))
- — secret validation record pre-distributed to the remote bank agent
- — one_way_ns = RTT/2: assumes symmetric forward/return paths; RTT includes end-host processing and queueing; NOT a physical lower bound and NOT a one-way measurement
- — clock error bound 1000 ns assumed (NTP-class)
$PY -m pytest tests/unit/networks/test_networks_ripe.py -q
RIPE Atlas msm 7714918 (RIPE Atlas REST API v2 (public measurement results)), retrieved 2026-10-11T21:15:00Z
Photon data · lane E
Error rates from the raw Schiansky 2023 archive
22 runs of real recorded photon coincidences (Zenodo 7979319), re-analysed independently. The paper's own estimator reproduces both published error rates.
H/V (Z basis)
- Pooled QBER
- 0.0143[0.01395, 0.01485] · 95% CIPHOTON-EXPERIMENT: Real recorded photon events with dataset hash
- Paper's estimator
- 0.01449[0.01433, 0.01465] · 95% CIPHOTON-EXPERIMENT: Real recorded photon events with dataset hash
- Published
- 0.0145[0.0144, 0.0146]PUBLISHED-AGGREGATE: Number quoted from a paper table or text
- run QBER, 95% Clopper–Pearson
- published ±
- pooled hull
| Run | Errors | Sifted | QBER | 95% CI |
|---|---|---|---|---|
| linear/chunk00 | 2839 | 194044 | 0.01463 | 0.01410–0.01517 |
| linear/chunk01 | 2957 | 203065 | 0.01456 | 0.01405–0.01509 |
| linear/chunk02 | 2838 | 201576 | 0.01408 | 0.01357–0.01460 |
| linear/chunk03 | 2896 | 203339 | 0.01424 | 0.01373–0.01477 |
| linear/chunk04 | 2766 | 200220 | 0.01381 | 0.01331–0.01434 |
| linear/chunk05 | 2785 | 202433 | 0.01376 | 0.01325–0.01427 |
| linear/chunk06 | 2775 | 163410 | 0.01698 | 0.01636–0.01762 |
| linear/chunk07 | 2786 | 203927 | 0.01366 | 0.01316–0.01417 |
| linear/chunk08 | 2910 | 203179 | 0.01432 | 0.01381–0.01485 |
| linear/chunk09 | 2850 | 204412 | 0.01394 | 0.01344–0.01446 |
| linear/chunk10 | 2829 | 204217 | 0.01385 | 0.01335–0.01437 |
Runs are overdispersed (χ²/dof = 10.5): per-run Clopper–Pearson intervals assume i.i.d. trials, so the pooled hull is the conservative interval. This is run-to-run heterogeneity, not temporal drift: a run-level Welch test of first vs second half gives p = 0.59 (no evidence of drift). Calibration record schiansky2023-zenodo7979319-Z-v1.
D/A (X basis)
- Pooled QBER
- 0.03287[0.03232, 0.03359] · 95% CIPHOTON-EXPERIMENT: Real recorded photon events with dataset hash
- Paper's estimator
- 0.03279[0.03256, 0.03302] · 95% CIPHOTON-EXPERIMENT: Real recorded photon events with dataset hash
- Published
- 0.0328[0.0327, 0.0329]PUBLISHED-AGGREGATE: Number quoted from a paper table or text
- run QBER, 95% Clopper–Pearson
- published ±
- pooled hull
| Run | Errors | Sifted | QBER | 95% CI |
|---|---|---|---|---|
| diagonal/chunk00 | 7192 | 220008 | 0.03269 | 0.03195–0.03344 |
| diagonal/chunk01 | 6950 | 208581 | 0.03332 | 0.03255–0.03410 |
| diagonal/chunk02 | 6935 | 216183 | 0.03208 | 0.03134–0.03283 |
| diagonal/chunk03 | 6771 | 209840 | 0.03227 | 0.03152–0.03303 |
| diagonal/chunk04 | 7142 | 215190 | 0.03319 | 0.03244–0.03395 |
| diagonal/chunk05 | 5841 | 160543 | 0.03638 | 0.03547–0.03731 |
| diagonal/chunk06 | 6975 | 216170 | 0.03227 | 0.03153–0.03302 |
| diagonal/chunk07 | 6936 | 217882 | 0.03183 | 0.03110–0.03258 |
| diagonal/chunk08 | 7083 | 218778 | 0.03238 | 0.03164–0.03313 |
| diagonal/chunk09 | 6930 | 216666 | 0.03198 | 0.03125–0.03273 |
| diagonal/chunk10 | 6662 | 194851 | 0.03419 | 0.03339–0.03501 |
Runs are overdispersed (χ²/dof = 9.9): per-run Clopper–Pearson intervals assume i.i.d. trials, so the pooled hull is the conservative interval. This is run-to-run heterogeneity, not temporal drift: a run-level Welch test of first vs second half gives p = 0.33 (no evidence of drift). Calibration record schiansky2023-zenodo7979319-X-v1.
- Not observable from the archive
- detection efficiency, channel loss, multiphoton probability, singles rates / accidental-coincidence rate, absolute run times — left null, never filled from the paper.
- Dataset
- Zenodo 7979319 · CC-BY-4.0 · sha256 ef73667f…dc800130
- Reproduce
- $PY -m qmint.photonics.schiansky2023
Reproducibility
Regenerate every number on this page
Commands (repo root, $PY = shared venv python)
$PY -m qmint.classical.run_baselines --pqc-reps 5$PY -m qmint.photonics.schiansky2023$PY -m pytest tests/unit/networks -q$PY -m qmint.proofs.reproduce
- Lane H records generated at
- 02aeab190025e51992cca0736f86dc731c1abee2
- Environment lock
- requirements.lock sha256:5a34f779b7d112e659a10d5c3aa8cb87793fde49b4f6adc0c68cd6f8e024f3e8
- Evidence files
- reports/classical_baseline_notes/benchmark_records.jsonlreports/classical_baseline_notes/comparability_matrix.csvreports/data_working/schiansky2023/per_run.csvreports/data_working/schiansky2023/summary.jsonreports/data_working/schiansky2023/calibration_records.json