Pith. sign in

REVIEW 4 major objections 4 minor 14 references

This paper claims that a per-interval leakage estimator with a dual-threshold kill-switch can enforce operational privacy for multi-tenant quantum cloud jobs, aborting any job whose observable timing and metadata diverge too far from a lock

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

NADGO combines dummy-gate padding, timing randomization, and routing with a per-interval leakage monitor to enforce operational privacy budgets for quantum cloud workloads.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection Integration is real and the engineering is honest, but Lemma IV.1 is unproven and false as stated, so the security guarantee does not hold. the 4 major comments →

arxiv 2509.00812 v1 pith:OJAD423A submitted 2025-08-31 cs.CR

Adaptive t Design Dummy-Gate Obfuscation for Cryogenic Scale Enforcement

Samuel Punch , Krishnendu Guha This is my paper
classification cs.CR
keywords quantum cloud securityoperational side channelst-design paddingleakage estimationper-interval policymulti-tenant quantum computingcryo-CMOS controlobfuscation
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Multi-tenant quantum clouds leak circuit structure through queue timing, scheduler metadata, and co-tenant interference, even when computation outputs are correct. The paper introduces NADGO, a scheduling and obfuscation stack that caps this operational leakage interval by interval, using t-design dummy-gate padding, particle-filter timing jitter, topology-aware subcircuit routing, and an online monitor that compares observed execution cadence against a fixed design-only reference. If the estimated leakage crosses a high threshold, the job is aborted and the decision is appended to a hash-chained audit log. A proof sketch argues that, over admitted intervals, any polynomial-time adversary's advantage is bounded by the per-interval budget plus calibration and synchronization errors. The stated payoff is a practical, auditable operational-privacy guarantee at microsecond control-loop speed with lower latency and cryogenic power than static padding.

Core claim

The central claim is that operational privacy can be enforced per interval, not just averaged over a job. The estimator bDelta_t = min{KL(p_hat_t || p_des_t) + beta_est ||Lambda_C||_F, Delta_kill} compares the empirical cadence distribution of observable features with a locked distribution collected from design-only executions, adds a crosstalk penalty, and caps the value at the kill threshold. Lemma IV.1 states that, under assumptions of feature sufficiency, Lipschitz crosstalk impact, and conditional independence, this estimator upper-bounds the true instantaneous leakage sup_A I(U_real; A(S_t)|Lambda_C, Y). Theorem V.1 then bounds the OP-IND advantage of any probabilistic polynomial-time

What carries the argument

The per-interval leakage estimator with locked calibration artifacts. It maps a fixed 2560-codeword codebook built from four observable features (inter-dispatch gap, batch-size ratio, queue state, telemetry proxy) into empirical histograms, then measures KL divergence to a locked design-only reference plus a crosstalk penalty. This single quantity is simultaneously the policy signal, the routing signal, and the audited evidence, so the entire enforcement argument rests on how faithfully it tracks the true conditional mutual information.

Load-bearing premise

The guarantee holds only if the finite feature alphabet used by the monitor captures every observable an adversary can exploit, and if the KL divergence to the design-only reference tracks the true mutual information within a known calibration error.

What would settle it

Run the same policy on a device with more than four qubits while an adversary records a channel outside the monitored features, such as supply-current fluctuations, qubit readout rise times, or co-tenant crosstalk patterns, and check whether circuit-depth classification still succeeds even though bDelta_t stays near zero.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • If correct, cloud providers can offer operational privacy as a measurable service-level objective: nominal jobs run with interval-abort rates below 1 percent, while adversarial timing or batch-size modulation reliably triggers concentrated aborts.
  • Burst leakage is caught rather than hidden in averages: because the policy operates per interval, a short spike that would disappear in a job-level aggregate pushes bDelta_t toward Delta_kill and aborts execution.
  • Heterogeneous backends can be used without abandoning privacy: CASQUE routing selects a backend by a cost that includes a leakage-risk term fed by the same estimator used for enforcement.
  • The append-only, hash-chained audit log makes compliance checkable after the fact, supporting fair multi-tenancy and operator accountability.
  • At matched leakage targets, microbenchmarks indicate lower latency and cryogenic power than static padding, making the privacy guarantee cheaper than the standard timing-regularization baseline.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The enforceability guarantee is only as broad as the monitored feature alphabet. If an adversary exploits an observable outside S_t, such as supply-current fluctuations, readout rise times, or pulse-level details, the estimator could read near zero while real leakage continues; testing this would require a channel beyond the four monitored features.
  • The same estimator-and-abort pattern is not inherently quantum: the t-design padding and routing are quantum-specific, but the KL-versus-locked-reference monitor with a dual-threshold kill-switch could transfer to other multi-tenant compute settings where job structure leaks through cadence.
  • The one-shot quantile-alignment transfer of thresholds from Tier I to Tier II is a reproducibility point: a third environment with different drift dynamics would reveal whether the fixed calibration remains valid or needs re-alignment.
  • The proof is a proof sketch built on calibration assumptions; a concrete check would be measuring |E[bDelta_t] - Delta_I_t| on a larger, drifting device and verifying whether the claimed per-interval bound still holds with the stated epsilon_est.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The paper proposes NADGO, a layered obfuscation and scheduling stack for multi-tenant quantum cloud services. The claimed contribution is per-interval enforcement of an information-theoretic operational-privacy bound: a real-time estimator bDelta_t, defined as the KL divergence between the empirical cadence distribution and a locked design-only reference plus a crosstalk penalty, is used to abort jobs that exceed a threshold (Delta_budget, Delta_kill). The security argument rests on Lemma IV.1, which states that the KL-based estimator upper-bounds the true instantaneous leakage Delta_I_t = sup_A I(U_real; A(S_t)|Lambda_C,Y), and on Theorem V.1, which converts this into an OP-IND advantage bound. The experimental sections report Monte Carlo and hardware-emulation results for attack detection, abort rates, latency, and power. Section VII, however, defers 'succinct proofs' of interval-level policy enforcement to future work, and no proof of the central lemma is given in the text.

Significance. If the central bound were valid, NADGO would be an attractive integrated defense for quantum-cloud operational side channels, combining t-design padding, particle-filter timing randomization, topology-aware routing, a 6.3-microsecond monitor, and an auditable kill-switch. The engineering scaffold—locked calibration artifacts, hash-chained audit logging, a hardware prototype, and per-interval instrumentation—is concrete and potentially useful. However, the core privacy guarantee is not established: the load-bearing Lemma IV.1 is unproven and, as stated, false in general, and the evaluation measures only the design-dependent quantity bDelta_t rather than the information-theoretic leakage it is supposed to enforce. The paper currently reads as an anomaly-detection system with a privacy-theoretic framing, not as a validated operational-privacy mechanism.

major comments (4)
  1. [§IV-A(b), Lemma IV.1] Lemma IV.1 is the bridge from the computable estimator to the true leakage Delta_I_t, but it is stated without proof. The asserted inequality sup_A I(U_real; A(S_t)|Lambda_C,Y) <= D_KL(bp_t || p_des_t) + beta_est||Lambda_C||_F does not follow from (A1)-(A3). It is also false as stated: KL divergence between the marginal of S_t under the real job and under the design reference can be zero even when S_t is a deterministic function of U_real with the same quantized marginal, e.g., S_t = g(U_real) with P_real(g) = P_des(g). Then the left side is H(U_real) while the right side reduces to the crosstalk penalty. Since Theorem V.1 inherits this step, the claimed OP-IND bound is unsubstantiated. The paper itself acknowledges in Section VII that 'succinct proofs of interval-level policy enforcement' are future work.
  2. [§V, Theorem V.1] The theorem assumes two unmeasured quantities: calibration error |E[bDelta_t] - Delta_I_t| <= eps_est(t) and bounded sync error eps_sync. No method is given to bound or estimate either, and the experiments never measure Delta_I_t directly. The statement that 'calibration |E[bDelta_t] - Delta_I_t| <= eps_est' is therefore an unsupported assumption, not a verified hypothesis. Additionally, Section V says the OP-IND game is 'defined above', but no formal game definition appears anywhere in the manuscript; without it, the advantage expression in Eq. (4) is not well-defined.
  3. [§VI, Figs. 3-6] The empirical evaluation is substantially self-referential. bDelta_t is defined as KL divergence to a design-only reference; the padding and timing layers are designed to keep bDelta_t small; thresholds are fit to baseline quantiles of bDelta_t; and 'leakage within budget' is reported for bDelta_t. The AUC/abort-rate results therefore show that the estimator responds to anomalies in the four chosen cadence features (Delta_gap, b, q, zeta), not that it controls or estimates information leakage about the hidden computation U_real. In particular, the statement that NADGO 'maintains leakage within budget' is a property of the estimator relative to its own reference, not an independent measure of privacy.
  4. [§IV-A(b) and §V] Theorem V.1's proof idea invokes a hybrid over admitted intervals and Pinsker's inequality, which requires per-interval independence or a suitable martingale structure. Assumption (A3) asserts conditional independence across intervals given (Lambda_C,Y), but the orchestration itself is stateful: the particle filter carries particles and queue state across intervals, and CASQUE routing depends on past bDelta_t values. These dependencies are not accounted for, so the product-distribution argument is not justified even if Lemma IV.1 were repaired. The 'mild regularity of Lambda_C' in Lemma IV.1 is also never specified.
minor comments (4)
  1. [Figures 3-6] The rendered text contains long literal '/uni000000...' sequences instead of figure content. The PDF figure insertion is broken and must be regenerated before review.
  2. [Eq. (2)] The estimator bDelta_t is capped at Delta_kill, so the notation 'min(..., Delta_kill)' makes the behavior at the threshold clearer if the abort condition is '>=' rather than '>'.
  3. [§III] The threat model excludes telemetry-side channels from experimental scope, but the cadence feature set includes a telemetry proxy zeta. The relationship between the excluded and included telemetry should be clarified.
  4. [§VI-B] Reporting latency and power per admitted interval can hide denial-of-service costs from aborted episodes; a per-episode or total-cost view would give a more complete picture. The paper notes this only implicitly.

Circularity Check

1 steps flagged

Empirical 'leakage within budget' claim is self-fulfilling because thresholds are fit to baseline quantiles of the estimator; the formal bound is conditional on an unproven estimator lemma.

specific steps
  1. fitted input called prediction [Section VI, 'Thresholds (reporting & rationale)' and Section VI-A; Fig. 3]
    "We set thresholds per experiment (e.g., by job size n, workload, environment) by targeting baseline quantiles of Δ̂t with a small hysteresis gap, fixing a nominal false-alarm rate and stabilising decisions under distributional shift."

    The operational threshold Δbudget is defined as a baseline quantile of the estimator Δ̂t. Therefore the headline result 'NADGO maintains leakage within budget in nominal operation (interval-abort rate below 1 percent)' is guaranteed by construction: the budget was chosen so that a fixed fraction of baseline Δ̂t values fall below it. Reporting baseline mass 'concentrated below Δbudget' and a low abort rate is a restatement of the quantile choice, not independent evidence that Δ̂t tracks true information leakage ΔI_t. The attack-separation results are also computed on the same Δ̂t statistic, so they validate only the anomaly-detection behavior of the estimator, not the unproven Lemma IV.1 bridge between KL-divergence and mutual information.

full rationale

The paper's formal derivation (Theorem V.1) is conditional and not itself circular: it assumes a calibration error bound and uses Pinsker/hybrid arguments to convert per-interval budgets into an OP-IND advantage bound. However, the load-bearing link between the fitted KL statistic Δ̂t and true leakage ΔI_t is Lemma IV.1, which is asserted without proof; Section VII also lists 'succinct proofs of interval-level policy enforcement' as future work, confirming the proof gap. This is a correctness risk, not a circularity. The concrete circularity is in the empirical validation: thresholds are chosen per experiment as baseline quantiles of Δ̂t, so the claim that baseline operation stays within budget and aborts below 1% is a tautology. The attack-induced separation is not forced by construction, but it is measured on the same estimator used to set thresholds, so it does not independently establish that Δ̂t bounds information leakage. Because the central empirical claim reduces to the fitted threshold, and the formal guarantee rests on an unproven lemma, a score of 6 reflects partial circularity with independent mathematical content still possible if the lemma were supplied and proven.

Axiom & Free-Parameter Ledger

7 free parameters · 7 axioms · 0 invented entities

The central claim rests on assumptions (A1)-(A3) and an unproven calibration bound that connects bDelta_t to Delta_I_t, plus hand-chosen parameters for padding, timing, routing, and thresholds. The standard t-design and concentration results are external and acceptable. The absence of a measured epsilon_est and the threshold fitting on baseline data are the main burdens.

free parameters (7)
  • beta_est = not reported
    Crosstalk penalty weight in Eq (2); fixed but value not given.
  • Delta_budget, Delta_kill thresholds = per experiment, from baseline quantiles of bDelta
    Set per experiment by targeting baseline quantiles of the estimator; detection results depend on this fitting.
  • calibration windowing hyperparameters W, S, H, alpha, lambda = W=128, S=64, H=10, alpha=1/2, lambda=1e-3
    Chosen by hand for windowing and smoothing; no sensitivity analysis.
  • quantization codebook sizes = B_Delta=16, B_b=5, B_q=4, B_zeta=8
    Hand-chosen bins; the full codebook is locked but not published.
  • t-design parameters = t=ceil(log2 n), epsilon_des=0.02
    Approximation error and design order chosen ad hoc for padding.
  • particle filter parameters sigma_t, sigma_proc(q), ell_max, N = not reported
    Algorithm depends on these; no values are given.
  • CASQUE routing weights and hysteresis = not reported
    alpha, omega_leak, gamma, tau, m, delta are fixed across experiments but values are omitted.
axioms (7)
  • domain assumption Feature sufficiency (A1): the cadence feature alphabet S_t = (Delta_gap, b, q, zeta) captures all observable information relevant to U_obs.
    Stated in Section IV-A(b) as assumption (A1); not validated against actual adversarial observables.
  • domain assumption Lipschitz regularity of the crosstalk channel Lambda_C (A2).
    Assumed in Section IV-A(b) so the crosstalk penalty term can be added to the KL divergence.
  • domain assumption Conditional independence across intervals given (Lambda_C, Y) (A3).
    Assumed in Section IV-A(b); needed for per-interval composition in Theorem V.1.
  • ad hoc to paper Calibration accuracy: |E[bDelta_t] - Delta_I_t| <= epsilon_est(t) for all admitted intervals.
    Assumed in Theorem V.1; no measurement or proof is supplied, and it is exactly the bridge from the computable estimator to true leakage.
  • ad hoc to paper Bounded skew and jitter epsilon_sync.
    Assumed in Theorem V.1; no bound is derived or measured.
  • standard math t>=4 approximate designs are indistinguishable from Haar-random (Ambainis-Emerson 2007); local random circuits approximate designs (Brandao-Harrow-Horodecki 2016, Haferkamp 2022).
    Relied on for the security of dummy-gate padding; these are external published results.
  • standard math Pinsker's inequality and concentration bounds for empirical distributions (Mardia et al. 2019).
    Used in the proof sketch of Theorem V.1.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Adaptive t Design Dummy-Gate Obfuscation for Cryogenic Scale Enforcement." pith.science (2026). https://pith.science/paper/OJAD423A

@misc{pith2026250900812,
  author       = {Pith},
  title        = {Pith review of: Adaptive t Design Dummy-Gate Obfuscation for Cryogenic Scale Enforcement},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OJAD423A}},
  note         = {Machine review of arXiv:2509.00812}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

Cloud quantum services can reveal circuit structure and timing through scheduler metadata, latency patterns, and co-tenant interference. We introduce NADGO (Noise-Adaptive Dummy-Gate Obfuscation), a scheduling and obfuscation stack that enforces operational privacy for gate-model workloads by applying per-interval limits on observable information leakage. To support confidentiality and fair multi-tenancy, operators require a method to audit compliance at acceptable overheads. NADGO combines: (i) hardware-aware t-design padding for structured cover traffic, (ii) particle-filter timing randomization to mask queue patterns, (iii) CASQUE subcircuit routing across heterogeneous backends, and (iv) a per-interval leakage estimator with locked calibration artifacts and a dual-threshold kill-switch. We prototype the approach on a 4-qubit superconducting tile with cryo-CMOS control and evaluate both depth-varied local-random circuits and small QAOA instances. Monitoring runs at a 6.3 microsecond control interval, and per-interval decisions are recorded in an append-only, hash-chained audit log. Across Monte Carlo (Tier 1) and cloud-hardware emulation (Tier 2) evaluations, NADGO maintains leakage within budget in nominal operation (interval-abort rate below 1 percent) and under attack yields high separation with concentrated aborts. At matched leakage targets, microbenchmarks indicate lower latency and cryogenic power consumption than static padding, while end-to-end workloads maintain competitive cost envelopes.

Figures

Figures reproduced from arXiv: 2509.00812 by Krishnendu Guha, Samuel Punch.

Figure 1
Figure 1. Figure 1: Vertical system pipeline: policy-aligned compilation, secure padding, [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Threat model (evaluated scope): two parallel chains, A1 [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Tier I: leakage distributions. Histograms of Δˆ 𝑡 for baseline and adversarial workloads. Vertical lines denote the exact (Δbudget, Δkill) used in this experiment (from the code logs). Baseline mass lies below Δbudget; rl shifts the distribution; timing induces heavier right tails, approaching/exceeding Δkill. 70 75 80 85 90 95 100 _budget percentile (baseline 'none') 0 20 40 60 80 100 ABORT intervals (%) … view at source ↗
Figure 4
Figure 4. Figure 4: Tier I: threshold sensitivity. Varying Δbudget at a fixed hysteresis gap demonstrates robust detection across operating points: tighter budgets yield higher detection at the expense of higher abort rates, as expected. that approach or exceed Δkill. Vertical lines mark the exact (Δbudget, Δkill) emitted by the code/logs for each panel, making the operating point explicit relative to the induced tails. Sweep… view at source ↗
Figure 5
Figure 5. Figure 5: Tier II leakage histograms (per admitted interval). Left: baseline (none) mass lies below Δbudget. Middle: rl shifts Δˆ 𝑡 right. Right: timing induces heavier right tails that approach/exceed Δkill. Threshold lines are the exact per-experiment (Δbudget, Δkill) emitted by the code/logs. 4 8 Number of qubits (n) 0 10 20 30 40 ABORT intervals (%) 0.0% 0.0% 11.2% 2.6% 39.2% 10.4% Tier-2 (emulation): ABORT rate… view at source ↗
Figure 6
Figure 6. Figure 6: Tier II outcomes and costs (per admitted interval). Left: ABORT rate by attack and 𝑛 ∈ {4, 8}. Middle: latency vs. 𝑛 with 95% CIs. Right: power vs. 𝑛 with 95% CIs. Adversarial workloads (especially timing) raise ABORT incidence and increase latency/power, consistent with the heavier leakage tails in [PITH_FULL_IMAGE:figures/full_fig_p006_6.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

14 extracted references · 4 canonical work pages · 1 internal anchor

  1. [1]

    Private quantum computation: An introduction to blind quantum computing and related protocols,

    J. F. Fitzsimons, “Private quantum computation: An introduction to blind quantum computing and related protocols,”npj Quantum Information, vol. 3, no. 1, p. 23,

    work page

  2. [2]

    C. Lu, E. Telang, A. Aysu, and K. Basu,Quantum leak: Timing side-channel attacks on cloud-based quantum services, 2024. arXiv: 2401.01521[cs.ET]. [Online]. Available: https://arxiv.org/abs/2401.01521

    Pith/arXiv arXiv 2024

  3. [3]

    Side-channel leakage in sfq circuits and related attacks on qubit control and readout systems,

    Y. Mustafa and S. Köse, “Side-channel leakage in sfq circuits and related attacks on qubit control and readout systems,” IEEE Transactions on Applied Superconduc- tivity, vol. 33, no. 6, pp. 1–7, 2023.doi: 10.1109/TASC. 2023.3277864. [Online]. Available: https://doi.org/10. 1109/TASC.2023.3277864

    arXiv 2023

  4. [4]

    Ambainis and J

    A. Ambainis and J. Emerson,Quantum 𝑡-designs: 𝑡- wise independence in the quantum world, 2007. arXiv: quant-ph/0701126 [quant-ph]. [Online]. Available: https://arxiv.org/abs/quant-ph/0701126

    Pith/arXiv arXiv 2007

  5. [5]

    Local random quantum circuits are approximate polynomial-designs,

    F. G. S. L. Brandão, A. W. Harrow, and M. Horodecki, “Local random quantum circuits are approximate polynomial-designs,” Communications in Mathematical Physics, vol. 346, no. 2, pp. 397–434, Aug. 2016.doi: 10.1007/s00220-016-2706-8. [Online]. Available: https: //doi.org/10.1007/s00220-016-2706-8

    doi:10.1007/s00220-016-2706-8 2016

  6. [6]

    Random quantum circuits are approxi- mate unitary𝑡-designs in depth𝑂(𝑛𝑡5+𝑜(1)),

    J. Haferkamp, “Random quantum circuits are approxi- mate unitary𝑡-designs in depth𝑂(𝑛𝑡5+𝑜(1)),” Quantum, vol. 6, p. 795, 2022.doi: 10.22331/q-2022-09-08-795. arXiv: 2203.16571 [quant-ph]. [Online]. Available: https://doi.org/10.22331/q-2022-09-08-795

    Pith/arXiv arXiv 2022

  7. [7]

    On the use of calibration data in error-aware compilation techniques for nisq devices,

    H. Kurniawan, L. Rodríguez-Soriano, D. Cuomo, C. G. Almudéver, and F. G. Herrero, “On the use of calibration data in error-aware compilation techniques for nisq devices,” in 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), vol. 01, 2024, pp. 338–348.doi: 10.1109/QCE60285.2024.00048. [Online]. Available: https://doi.org/10.110...

    arXiv 2024

  8. [8]

    Noise-Adaptive Quantum Compilation Strategies Evaluated with Application-Motivated Benchmarks

    D. Ferrari and M. Amoretti, “Noise-adaptive quantum compilation strategies evaluated with application-motivated benchmarks,” arXiv preprint arXiv:2108.11874, 2021. doi: 10.48550/arXiv.2108. 11874. arXiv: 2108 . 11874 [quant-ph]. [Online]. Available: https://doi.org/10.48550/arXiv.2108.11874

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2108.11874 2021

  9. [9]

    Mardia, J

    J. Mardia, J. Jiao, E. Tánczos, R. D. Nowak, and T. Weissman, Concentration inequalities for the em- pirical distribution, 2019. arXiv: 1809.06522[cs.IT]. [Online]. Available: https://arxiv.org/abs/1809.06522

    Pith/arXiv arXiv 2019

  10. [10]

    Quantumnas: Noise-adaptive search for robust quantum circuits,

    H. Wanget al., “Quantumnas: Noise-adaptive search for robust quantum circuits,” in2022 IEEE International Symposium on High-Performance Computer Architecture (HPCA), IEEE, Apr. 2022, pp. 692–708.doi: 10.1109/ HPCA53966.2022.00057. [Online]. Available: https: //doi.org/10.1109/HPCA53966.2022.00057

    arXiv 2022

  11. [11]

    Cryogenic measurement of cmos devices for quantum technologies,

    J. Pérez-Bailón, M. Tarancón, S. Celma, and C. Sánchez- Azqueta, “Cryogenic measurement of cmos devices for quantum technologies,” IEEE Transactions on Instru- mentation and Measurement, vol. 72, pp. 1–7, 2023. doi: 10.1109/TIM.2023.3325446. [Online]. Available: https://doi.org/10.1109/TIM.2023.3325446

    work page arXiv 2023

  12. [12]

    Method for efficient large- scale cryogenic characterization of cmos technologies,

    J. Eastoe, G. M. Noah, D. Dutta, A. Rossi, J. D. Fletcher, and A. Gomez-Saiz, “Method for efficient large- scale cryogenic characterization of cmos technologies,” IEEE Transactions on Instrumentation and Measurement, vol. 74, pp. 1–10, 2025

    work page 2025

  13. [13]

    Opaque: Obfuscating phase in quantum circuit compilation for efficient ip protection,

    A. Rehman, V. Langford, J. John, and Y. Liu, “Opaque: Obfuscating phase in quantum circuit compilation for efficient ip protection,” in 2025 26th International Symposium on Quality Electronic Design (ISQED) , 2025, pp. 1–6. doi: 10.1109/ ISQED65160.2025. 11014313. [Online]. Available: https://doi.org/10.1109/ ISQED65160.2025.11014313

    arXiv 2025

  14. [2017]

    [Online]

    doi: 10.1038/s41534-017-0025-3. [Online]. Available: https://doi.org/10.1038/s41534-017-0025-3

    doi:10.1038/s41534-017-0025-3

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.