Pith. sign in

REVIEW 2 major objections 6 minor 21 references

Perturbation Analysis of Maximal Quantum Leakage

T0 review · 2 major / 6 minor · reviewed 2026-08-02 · deepseek-v4-flash

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

Pith's one-line read Small perturbations of quantum encodings can change maximal quantum leakage only linearly.

desk verdict The continuity bound is plausible and the question is well placed, but Theorem 1's proof relies on a false rank bound for a sum of non-orthogonal projectors, so the central result is not established as written. read the letter →

arxiv 2607.14469 v1 pith:B7MJHDFD submitted 2026-07-16 quant-ph cs.ITmath.IT

Zijia Zhao , Shuixin Xiao , Farhad Farokhi This is my paper · ORCID
classification quant-phcs.ITmath.IT MSC 81P4594A17
keywords maximalquantumleakageencodingperturbationtracedistancecontinuityboundworst-caseguessingadvantagePOVMfidelityrelativeentropystabilityanalysis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This paper asks how a worst-case quantum information-leakage measure — maximal quantum leakage — responds when the quantum states encoding a secret are slightly wrong. Its central claim is a tight continuity bound: if each ideal state differs from its perturbed version by at most ε in trace distance, then the adversary's multiplicative guessing advantage can change by no more than (ε/2)·min{|X|,d}, where |X| is the alphabet size and d the Hilbert-space dimension. The bound is shown to be attainable by an explicit pair of ensembles, and the same argument yields looser sufficient conditions based on quantum fidelity and quantum relative entropy. Establishing such stability matters because real quantum devices never prepare states exactly as intended, so leakage guarantees must tolerate implementation errors.

What carries the argument

The engine of the proof is the identity 2^{Q_ρ} = sup_{POVM {F_y}} Σ_y max_x Tr(ρ_x F_y), which converts the leakage measure into an optimization over measurements. Perturbation is then analyzed via the Jordan–Hahn decomposition of each Hermitian difference ρ_x − ρ'_x = Δ_x^+ − Δ_x^−, with Tr(Δ_x^+) = Tr(Δ_x^−) ≤ ε/2. The step that produces the clean linear constant is the assertion that the projectors P_x onto the positive eigenspaces of the differences satisfy Tr(Σ_x P_x) ≤ min{|X|,d}; this is the load-bearing step that fixes the bound's dependence on the alphabet size and dimension.

What would settle it

Take d=3, |X|=2, and the four diagonal states ρ1=diag(0.4,0.4,0.2), ρ1'=diag(0.2,0.2,0.6), ρ2=diag(0.4,0.2,0.4), ρ2'=diag(0.2,0.6,0.2), each pair at trace distance 0.8. The positive-eigenspace projectors are P1=diag(1,1,0) and P2=diag(1,0,1), so Tr(P1+P2)=4 > 2 = min{|X|,d}; this single calculation disproves the proof's key lemma, so any valid proof of the claimed bound must replace that step.

Watch

Extended reading notes

Core claim

In the paper's own terms, the discovery is that maximal quantum leakage is stable under small perturbations of the encoding: for two ensembles with the same alphabet and Hilbert space satisfying ||ρ_x − ρ'_x||_1 ≤ ε for every x, the variation of the leakage exponent satisfies |2^{Q_ρ} − 2^{Q_ρ'}| ≤ (ε/2)·min{|X|,d}. Here 2^{Q_ρ} is the optimal worst-case multiplicative increase in guessing probability, computed as the supremum over POVMs of Σ_y max_x Tr(ρ_x F_y). The proof decomposes each difference ρ_x − ρ'_x into positive and negative parts and bounds the contribution of the positive parts; an explicit construction with states on a d-dimensional space attains equality, indicating that the

Load-bearing premise

The proof's central constant rests on the assertion that the sum of the positive-eigenspace projectors of the state differences has trace at most min{|X|,d}; this assertion is not true in general, so the theorem is not established as written.

Editorial extensions

If this is right

  • If Theorem 1 is correct, a quantum device whose state preparations are ε-accurate in trace distance can change an adversary's multiplicative guessing advantage by at most (ε/2)·min{|X|,d}.
  • The saturation example implies that the dependence on |X| and d in the bound cannot be improved in general.
  • The fidelity and relative-entropy corollaries give easy-to-check sufficient conditions: squared fidelity above 1−ε or relative entropy below ε yield leakage changes bounded by √ε·min{|X|,d} and √(2 ln 2 ε)/2 · min{|X|,d}, respectively.
  • The numerical experiments indicate that the fidelity- and relative-entropy-based sufficient conditions can be loose, so trace-distance-based control is preferable when available.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The projector-trace assertion used to obtain the min{|X|,d} constant is not valid when the positive eigenspaces of the state differences overlap; a corrected proof or a counterexample would determine whether the stated constant survives or needs a geometry-dependent correction.
  • One can test the claimed bound numerically by sampling random ensembles whose state differences have heavily overlapping positive eigenspaces and computing the leakage exponent; any pair violating the bound would force a restatement of the theorem.
  • The same perturbation framework could be adapted to neighbouring operational leakage measures, such as pointwise maximal leakage or Rényi variants, where the worst-case guessing advantage takes a different functional form.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 6 minor

Summary. The paper studies the robustness of maximal quantum leakage (MQL) under perturbations of the quantum encoding. Its main result, Theorem 1, claims a Lipschitz-type continuity bound: if each state in an ensemble is perturbed by at most ε in trace distance, then the change in 2^{Q_ρ} is bounded by (ε/2)·min{|X|,d}. The authors provide an example that attains the bound in the regime |X| ≤ d, and derive corollary bounds in terms of quantum fidelity and quantum relative entropy via standard inequalities. Numerical experiments are used to illustrate that the fidelity- and relative-entropy-based bounds are loose in practice. The paper is clearly written and the result, if properly established, is a clean stability statement for an operational quantum leakage measure.

Significance. If Theorem 1 is correctly established, it provides a simple, interpretable Lipschitz constant for MQL as a function of the encoding states, depending only on alphabet size and Hilbert-space dimension. The tightness example is elegant and correctly saturates the bound when |X| ≤ d. The corollaries are straightforward but connect the result to physically natural perturbation metrics. A strength of the paper is its transparency: it builds directly on the prior exact characterization of MQL in [12]. However, the proof of the main theorem contains a genuine mathematical error in the step that generates the min{|X|,d} constant. I verified that the theorem statement is nevertheless true, via a short alternative grouping argument, so the gap is repairable rather than fatal.

major comments (2)
  1. [Section III-A, proof of Theorem 1] The final step of the proof is invalid. The text asserts that Σ_x P_x has rank at most min{|X|,d} and concludes D ≤ (ε/2)min{|X|,d}. The rank bound is false: for d=3, |X|=2, P_1=|0⟩⟨0|+|1⟩⟨1| and P_2=|0⟩⟨0|+|2⟩⟨2|, rank(Σ_x P_x)=3>2. Moreover, even a valid rank bound would not control the trace, since Tr(Σ_x P_x)=Σ_x rank(P_x) (trace is 4 in the same example). The constant min{|X|,d} is produced by exactly this step, so the proof as written does not establish Theorem 1, and Corollaries 1–2 inherit the gap. I verified that the theorem is true: for fixed F, f(F)−g(F) ≤ Σ_y max_x Tr(A_x F_y) = Σ_x Tr(A_x G_x) with G_x = Σ_{y:x_y=x}F_y; then Tr(A_x G_x) ≤ Tr(A_x^+) ≤ ε/2 and also ≤ (ε/2)Tr(G_x), giving (ε/2)min{|X|,d}. Replacing the faulty block with this argument would repair the proof.
  2. [Section III-A, tightness example] The tightness example only treats |X| ≤ d, so it demonstrates attainability only when min{|X|,d} = |X|. Since Theorem 1 is stated for all |X| and d, the claim that the bound is tight in general requires a construction for |X|>d as well; such a construction exists (for instance, d=2, |X|=3 with ρ_x=I/2, ρ'_1=I/2+(ε/2)σ_z, ρ'_2=I/2−(ε/2)σ_z, ρ'_3=I/2, and the projective measurement in the eigenbasis of σ_z gives 2^{Q_ρ'}=1+ε). The authors should add an example for this regime or explicitly restrict the tightness statement.
minor comments (6)
  1. [Throughout] Equations are unnumbered, which makes precise cross-referencing difficult.
  2. [Section III-A, proof of Theorem 1] Typo: 'decomposition decomposition'. Also, the Jordan–Hahn decomposition is cited to a probability text [20]; a matrix-analysis reference would be more appropriate.
  3. [Theorem 1] The wording 'for any x ∈ X and ε > 0, if ||ρ_x−ρ'_x||_1 ≤ ε' should read 'if ||ρ_x−ρ'_x||_1 ≤ ε for all x ∈ X', since ε is a uniform perturbation radius.
  4. [Corollary 1] The phrasing 'some ε > 0' is unclear; suggest 'for some ε > 0 such that 1−ε < F(ρ_x,ρ'_x) ≤ 1 for all x'.
  5. [Section III-D] Figs. 1 and 2 appear as captions only; the numerical section should state the optimizer tolerance and sample sizes so the reported maximum deviations are reproducible.
  6. [Section II-C] The MQL characterization and the bound 2^{Q_ρ} ≤ |X| are imported from [12]; these should be listed as explicit preliminary facts for self-containedness.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the continuity bound is an independent perturbation statement; self-citations to [12] supply the definition and characterization but are not used to prove the target result by construction.

full rationale

The paper's central claim, Theorem 1, is a Lipschitz-type bound on |2^Q_rho - 2^Q_rho'| in terms of trace distances ||rho_x - rho'_x||_1. This is not equivalent to its inputs: the MQL characterization 2^Q_rho = sup_{POVM} sum_y max_x Tr(rho_x F_y) is imported from the coauthor's prior work [12] as a definition/theorem, not fitted to the target data, and no fitted parameter is later relabeled as a prediction. The proof proceeds by bounding a difference of suprema through a fixed POVM and Jordan-Hahn decomposition; the conclusion does not reduce to the starting characterization. The tightness example uses Proposition 2 of [12] (2^Q_rho <= |X|) together with a direct POVM construction; again, the example is an application, not a circular re-derivation. The numerical section uses the algorithm from [12] only to compute MQL values, which does not make the analytical bounds outputs of that computation. There is no uniqueness theorem invoked to forbid alternatives, no ansatz smuggled in via citation, and no known result merely renamed. The manuscript does contain a serious mathematical gap: the assertion in Section III-A that 'the operator sum_x P_x has rank at most min{|X|, d}' is false for non-orthogonal projectors, and the trace bound is also uncontrolled; this invalidates Theorem 1 as written. But a false or unsupported step is a correctness flaw, not circularity: the step does not make the theorem an identity with its assumptions. Therefore the circularity score is 0, while the correctness risk is high and should be addressed separately.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No free parameters are fitted in the paper; the only constants come from standard inequalities. The central result depends on a prior exact characterization of MQL from a self-cited paper, plus standard quantum-information inequalities. The proof also implicitly depends on the false projector-trace assertion, which is a mathematical error rather than an admitted axiom and is therefore recorded as a red flag rather than here.

assumptions (5)
  • domain assumption 2^{Q_ρ} = sup_POVM Σ_y max_x Tr(ρ_x F_y) from [12, Theorem 1]
    The entire analysis operates on this characterization of maximal quantum leakage. It is a prior theorem by a coauthor, not re-derived here, and the paper's own proof starts from this expression.
  • standard math A POVM attaining the supremum in the MQL characterization exists
    Used at the beginning of Theorem 1's proof to fix a maximizing POVM. This follows from compactness of the POVM set and continuity, but is not explicitly justified.
  • standard math Jordan-Hahn decomposition of Hermitian operators into positive and negative parts
    Used in the proof of Theorem 1 to write ρ_x − ρ'_x = Δ_{x+} − Δ_{x−} and to relate traces to the trace norm.
  • standard math Fuchs–van de Graaf inequalities
    Used in Corollary 1 to convert fidelity constraints into trace-distance bounds. Standard, but the resulting bound is acknowledged to be loose.
  • standard math Quantum Pinsker inequality: ||ρ−σ||_1^2 ≤ 2 ln 2 · D(ρ||σ)
    Used in Corollary 2 to convert relative-entropy constraints into trace-distance bounds.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Perturbation Analysis of Maximal Quantum Leakage." pith.science (2026). https://pith.science/paper/B7MJHDFD

@misc{pith2026260714469,
  author       = {Pith},
  title        = {Pith review of: Perturbation Analysis of Maximal Quantum Leakage},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B7MJHDFD}},
  note         = {Machine review of arXiv:2607.14469}
}
read the original abstract

Maximal quantum leakage (MQL) is a worst-case information leakage measure that quantifies an adversary's inference advantage gained from accessing quantum encoding of classical data with arbitrary measurements. While MQL admits an exact characterization for a given ensemble of quantum states, its robustness to implementation imperfections has not been systematically studied. In this paper, we analyze the sensitivity of maximal quantum leakage under perturbations of the quantum encoding. We establish a continuity bound in terms of the trace distance between ideal and perturbed quantum states, and show, via an example, that this bound is attainable. We further derive fidelity-based and relative-entropy-based sufficient conditions for bounding the variation of maximal quantum leakage, and illustrate numerically that these conditions can be loose.

Figures

Figures reproduced from arXiv: 2607.14469 by the authors.

Figure 1
Figure 1. Maximum deviation of 2MQL under fidelity-based perturbations as a function of the radius ε [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Maximum deviation of 2MQL under relative-entropy-based pertur￾bations as a function of the radius ε. D. Numerical Illustration We provide numerical results to investigate the potential looseness of stability bounds based on both fidelity and quantum relative entropy. In our simulations, we fix the Hilbert space dimension to d = 3 and generate a large pool of random density operators sampled from the induced Ginibre … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 1 linked inside Pith

  1. [12]

    Maximal information leakage from quantum encoding of classical data,

    F. Farokhi, “Maximal information leakage from quantum encoding of classical data,”Physical Review A, vol. 109, no. 2, p. 022608, 2024

    2024

  2. [1]

    Privacy against statistical inference,

    F. du Pin Calmon and N. Fawaz, “Privacy against statistical inference,” inProceedings of the 2012 50th Annual Allerton Conference on Communication, Control, and Computing (Allerton), pp. 1401–1408, 2012

    2012

  3. [2]

    Information-theoretic approaches to privacy in estimation and con- trol,

    E. Nekouei, T. Tanaka, M. Skoglund, and K. H. Johansson, “Information-theoretic approaches to privacy in estimation and con- trol,”Annual Reviews in Control, vol. 47, pp. 412–422, 2019

    2019

  4. [3]

    Directed information and privacy loss in cloud-based control,

    T. Tanaka, M. Skoglund, H. Sandberg, and K. H. Johansson, “Directed information and privacy loss in cloud-based control,” inProceedings of the 2017 American control conference (ACC), pp. 1666–1672, 2017

    2017

  5. [4]

    Secure estimation and zero-error secrecy capacity,

    M. Wiese, T. J. Oechtering, K. H. Johansson, P. Papadimitratos, H. Sandberg, and M. Skoglund, “Secure estimation and zero-error secrecy capacity,”IEEE Transactions on Automatic Control, vol. 64, no. 3, pp. 1047–1062, 2018

    2018

  6. [5]

    Pointwise maximal leakage,

    S. Saeidian, G. Cervia, T. J. Oechtering, and M. Skoglund, “Pointwise maximal leakage,”IEEE Transactions on Information Theory, vol. 69, no. 12, pp. 8054–8080, 2023

    2023

  7. [6]

    An operational approach to information leakage,

    I. Issa, A. B. Wagner, and S. Kamath, “An operational approach to information leakage,”IEEE Transactions on Information Theory, vol. 66, no. 3, pp. 1625–1657, 2019

    2019

  8. [7]

    A case for maximal leakage as a side channel leakage metric,

    B. Wu, A. B. Wagner, and G. E. Suh, “A case for maximal leakage as a side channel leakage metric,”arXiv preprint arXiv:2004.08035, 2020

    arXiv 2004

Show all 21 references
  1. [8]

    Maximal leakage minimization for the Shannon cipher system,

    I. Issa, S. Kamath, and A. B. Wagner, “Maximal leakage minimization for the Shannon cipher system,” inProceedings of the 2016 IEEE International Symposium on Information Theory (ISIT), pp. 520–524, 2016

    2016

  2. [9]

    Optimal mechanisms under maximal leakage,

    B. Wu, A. B. Wagner, and G. E. Suh, “Optimal mechanisms under maximal leakage,” in2020 IEEE Conference on Communications and Network Security (CNS), pp. 1–6, IEEE, 2020

    2020

  3. [10]

    Quantum privacy and hypothesis-testing,

    F. Farokhi, “Quantum privacy and hypothesis-testing,” inProceedings of the 2023 62nd IEEE Conference on Decision and Control (CDC), pp. 2841–2846, 2023

    2023

  4. [11]

    Quantum differential privacy: An information theory perspective,

    C. Hirche, C. Rouz ´e, and D. S. Franc ¸a, “Quantum differential privacy: An information theory perspective,”IEEE Transactions on Information Theory, vol. 69, no. 9, pp. 5771–5787, 2023

    2023

  5. [13]

    Barycentric and pairwise R ´enyi quantum leakage with application to privacy-utility trade-off,

    F. Farokhi, “Barycentric and pairwise R ´enyi quantum leakage with application to privacy-utility trade-off,”Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences, vol. 480, no. 2303, 2024

    2024

  6. [14]

    Measuring quantum information leakage under detection threat,

    F. Farokhi and S. Kim, “Measuring quantum information leakage under detection threat,” inProceedings of the 2024 IEEE 63rd Con- ference on Decision and Control (CDC), pp. 781–786, 2024

    2024

  7. [15]

    M. M. Wilde,Quantum information theory. Cambridge, UK: Cam- bridge University Press, 2 ed., 2013

    2013

  8. [16]

    Discriminating states: The quantum Chernoff bound,

    K. M. Audenaert, J. Calsamiglia, R. Munoz-Tapia, E. Bagan, L. Masanes, A. Acin, and F. Verstraete, “Discriminating states: The quantum Chernoff bound,”Physical Review Letters, vol. 98, no. 16, p. 160501, 2007

    2007

  9. [17]

    Tight uniform continuity bounds for quantum entropies: Conditional entropy, relative entropy distance and energy constraints,

    A. Winter, “Tight uniform continuity bounds for quantum entropies: Conditional entropy, relative entropy distance and energy constraints,” Communications in Mathematical Physics, vol. 347, no. 1, pp. 291– 313, 2016

    2016

  10. [18]

    Continuity bounds for quantum entropies arising from a fundamental entropic inequality,

    K. Audenaert, B. Bergh, N. Datta, M. G. Jabbour, A. Capel, and P. Gondolf, “Continuity bounds for quantum entropies arising from a fundamental entropic inequality,”IEEE Transactions on Information Theory, vol. 71, no. 9, pp. 7029–7038, 2025

    2025

  11. [19]

    Continuity bounds on the quantum relative entropy,

    K. M. R. Audenaert and J. Eisert, “Continuity bounds on the quantum relative entropy,”Journal of Mathematical Physics, vol. 46, no. 10, p. 102104, 2005

    2005

  12. [20]

    Billingsley,Probability and Measure

    P. Billingsley,Probability and Measure. New York, NY , USA: John Wiley & Sons, 3 ed., 1995

    1995

  13. [21]

    Statistical ensembles of complex, quaternion, and real matrices,

    J. Ginibre, “Statistical ensembles of complex, quaternion, and real matrices,”Journal of Mathematical Physics, vol. 6, no. 3, pp. 440– 449, 1965

    1965

Pith tools

Reviewed August 2, 2026 · model on record in the stance chip above.