{"id":"76a693fc-5bb4-4b41-a05b-0b7b88bbc1fe","arxiv_id":"2411.14123","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Secure multi-terminal strong coordination over a multiple-access wiretap channel is characterized by new inner and outer bounds, with a tight region for conditionally independent sources over deterministic links and a demonstration that encoder cribbing strictly improves rates.","lead":"This paper derives rate regions for two agents coordinating random outputs over a shared noisy channel while keeping those outputs secret from an eavesdropper. It provides inner and outer bounds, a complete characterization in a special case, and shows that letting one agent see the other's channel input improves the achievable rates.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 3's converse does not verify that its auxiliary variables satisfy the product-form p.m.f. (9); under correlated W they generally fail, so the claimed tight characterization is not established.","rationale":"The reader's verdict focused on Theorem 4 being a proof sketch and on the restrictiveness of the Theorem 3 assumptions. My stress-test identifies a more load-bearing and previously unmentioned issue: the converse of Theorem 3 may not actually land inside the rate region it claims to characterize, because the auxiliary random variables constructed in Section VIII do not obviously satisfy the product-form joint distribution (9). This matters because Theorem 3 is the central tight result of the paper: if its converse is not valid, the complete characterization is unproven. The concern is concrete and verifiable: in the degenerate but permitted case X1 = X2 = W, the defined auxiliaries share an extensive common component (the future W symbols), so the conditional independence required by (9) is violated by a linear-in-n amount. A referee could settle this by examining whether the proof anywhere establishes the factorization p(u1,u2|x1,x2,w,t) = p(u1|x1,t)p(u2|x2,t) and p(x̃1,x̃2|t) = p(x̃1|t)p(x̃2|t); the current text does not. I therefore recommend conditional acceptance, with the condition being a complete and correct single-letterization argument for Theorem 3, rather than merely an expansion of the Theorem 4 Fourier-Motzkin elimination. I disagree with the reader's weakest_assumption because the decisive risk is not the restrictiveness of the assumptions but an apparent gap in the proof under those very assumptions.","tokens_in":32668,"tokens_out":33884,"duration_ms":295997,"concrete_test":"Set q(w) = Bern(1/2), X1 = X2 = W, legitimate channel Ỹ = (X̃1, X̃2), and any wiretap channel, e.g., Z̃ = (X̃1, X̃2 ⊕ N) with N ~ Bern(p). Take an arbitrary code satisfying the coordination and secrecy conditions, form U1i and U2i exactly as in Section VIII, and compute I(U1i; U2i | X1i, X2i, Wi). Since U1i contains W_{i+1..n} and U2i contains X_{2,i+1..n} = W_{i+1..n}, this conditional mutual information is Θ(n) > 0, contradicting the product-form requirement p(u1,u2|x1,x2,w,t) = p(u1|x1,t)p(u2|x2,t) in (9). If the authors intend a different auxiliary construction in the converse, they should exhibit it and verify the factorization; until then the single-letterization is incomplete.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The converse proof of Theorem 3 (Section VIII) derives the single-letter inequalities using U1i = (K1, X_{1,i+1}^n, Ỹ_1^n, W_{~i}) and U2i = (K2, X_{2,i+1}^n, Ỹ_2^n), then relies on Appendix B for cardinality bounds. However, Theorem 3's rate region is quantified over p.m.f.s of the product form (9), which requires p(u1,u2|x1,x2,w,t) = p(u1|x1,t)p(u2|x2,t) and p(x̃1,x̃2|t) = p(x̃1|t)p(x̃2|t). The proof never checks that the auxiliaries extracted from an arbitrary code satisfy this structure. In general they do not: because q satisfies I(X1;X2|W)=0 but W may still be correlated with both sources, U1i contains W_{~i} while U2i contains X_{2,i+1..n}; for j > i, W_j and X_{2,j} are dependent even after conditioning on (X1i, X2i, Wi). For example, if X1 = X2 = W with W uniform Bernoulli, then both auxiliaries contain the future symbols W_{i+1..n}, giving I(U1i; U2i | X1i, X2i, Wi) = Θ(n) > 0. Thus the constructed pair does not factor as required by (9), and the derived bounds may hold only for an auxiliary distribution outside the stated region. This is a gap in the proof of the paper's headline result, independent of the Theorem 4 proof-outline issue.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies strong coordination with secrecy over a discrete memoryless multiple-access wiretap channel (MAC-WT). Two encoders observe i.i.d. correlated sources (X1^n, X2^n) and share pairwise independent randomness (K1, K2) with the decoder; the decoder also observes side information W^n and the legitimate channel output Ytilde^n, and must emit Y^n so that (X1^n, X2^n, W^n, Y^n, Ztilde^n) is close in total variation to q_{X1X2WY}^{(n)} times p_{Ztilde^n}. The authors derive an inner bound (Theorem 1) and an outer bound (Theorem 2), a claimed complete characterization for conditionally independent sources given W and a deterministic legitimate channel with Ytilde = (f1(Xtilde1), f2(Xtilde2)) under unlimited R02 (Theorem 3), and an inner bound for non-causal cribbing at Encoder 1 (Theorem 4). Section V gives an example where cribbing improves the achievable rates. Proofs use OSRB, Slepian-Wolf decoding, the functional representation lemma, and perturbation cardinality bounds.","tokens_in":32985,"tokens_out":16057,"duration_ms":154089,"significance":"If the Theorem 3 characterization is correct, the paper contributes a tight multi-terminal secure strong-coordination result over a wiretap channel, a regime where matching inner and outer bounds are rare; the example and its Proposition 1 converse are also instructive. The general inner/outer bounds appear to follow standard and identifiable techniques, and I did not find obvious errors in the proofs of Theorems 1 and 2 or in Proposition 1. The value of the manuscript is, however, conditional on Theorem 3: the converse currently does not show that the auxiliary variables live in the product-form family (9) over which the theorem is quantified, and Theorem 4 lacks a complete proof. Neither issue is cosmetic.","major_comments":[{"comment":"The converse of Theorem 3 does not establish that its auxiliary variables satisfy the product-form p.m.f. on which the theorem's rate region is defined. The proof sets U1i = (K1, X_{1,i+1}^n, Ytilde_1^n, W_{~i}) and U2i = (K2, X_{2,i+1}^n, Ytilde_2^n), and then uses these in I(U1; X1|W,T), I(U2; X2|W,T), and I(U1; X1,Y|X2,W,T). Region (9) requires p(u1,u2|x1,x2,w,t) = p(u1|x1,t)p(u2|x2,t), equivalently U1 indep U2 given (X1,X2,W,T) together with the conditional-independence structure inherited from the sources. The constructed variables generally fail this: for example, take q with X1 = X2 = W, W uniform on {0,1}; then I(X1;X2|W)=0, but U1i contains W_{~i} and U2i contains X_{2,i+1}^n, so both contain W_{i+1},...,W_n and I(U1i;U2i|X1i,X2i,Wi) = Theta(n). Thus the derived single-letter inequalities may describe a point outside the region in Theorem 3. The passage from time-sharing averages over T to a product-form distribution is not a routine step here and is not provided. This gap affects the paper's headline characterization; Appendix B's cardinality bounds do not repair it, because they are derived starting from a p.m.f. already assumed to have the form (81)/(9).","section":"VIII, Eqs. (8)-(9) and (65)-(66)"},{"comment":"The proof of Theorem 4 is an outline rather than a proof. After writing the joint distribution (67) and the two binning conditions (68)-(69), the text states that the Slepian-Wolf constraints, elimination of (F1,F2), strong coordination, and secrecy follow similarly to Theorem 1, and that Fourier-Motzkin elimination gives (10a)-(10h). Because cribbing changes the way Encoder 1 generates (U1,V1) and the channel input, the OSRB equivalence, the Slepian-Wolf rate conditions, and the secrecy analysis all need to be verified with the new conditional structure; this is not shown. As Section V's improvement claim relies on Theorem 4, the example currently inherits this missing support.","section":"IX, Theorem 4"}],"minor_comments":[{"comment":"The paragraph after (1) says that strong secrecy holds 'provided that the total variation distance goes to zero exponentially in n', but Definition 2 only requires the limit in (1) to vanish; the definition and the surrounding text should be reconciled.","section":"Definition 2"},{"comment":"The notation R_{secrecy_noisy-coord, R02 -> infinity} denotes a projection with 'exists R02', not an actual limit; the notation may mislead a reader into thinking a limiting construction is being used.","section":"Definition 3"},{"comment":"In the p.m.f. (11), the roles of V1 and V2 in the cribbing scheme are less transparent than in Theorem 1; a short explanation of why the cribbing structure changes the wiretap-coding auxiliary variables in this way would improve readability.","section":"Theorem 4 statement"}],"recommendation":"major_revision","confidential_remarks":"The decisive issue for the editor is the converse of Theorem 3: the product-form gap is central and not a routine detail. I am recommending major revision rather than rejection because the general framework and the special-case statement may well be repairable, but the paper as written does not prove its headline tight characterization. I would also ask for a complete proof of Theorem 4 before relying on the cribbing conclusions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a serious piece of information-theoretic work: it formulates secure multi-terminal strong coordination over a MAC-WT and gives an OSRB-based achievable region, an outer bound, and a cribbing extension. The inner bound proof is detailed and the cribbing example is genuinely nice—it shows a strict improvement and its converse is worked out. If you work on coordination or wiretap coding, the general bounds are worth having.\n\nBut the spotlight result, Theorem 3, has a load-bearing gap. The converse constructs U1i and U2i that include future source and side-information symbols: U1i contains W_{~i}, U2i contains X_{2,i+1..n}. The rate region is defined over p.m.f.s of the product form p(u1|x1,t)p(u2|x2,t), which requires U1 and U2 to be conditionally independent given the sources and time-sharing. The converse never checks this. Under the theorem's own assumptions it can fail badly: take X1 = X2 = W with W uniform Bernoulli. Then I(X1;X2|W)=0, yet after conditioning on (X1i,X2i,Wi) both auxiliaries contain the same future W's, giving I(U1i;U2i|...) = Θ(n). So the constructed pair is nowhere near product form, and the derived inequalities do not establish membership in the stated region. This is not a technicality; it undermines the claimed complete characterization.\n\nThe rest is more defensible. The Theorem 4 proof is admittedly an outline—Fourier–Motzkin elimination is asserted, not displayed—but that's an inner bound only, so the gap is less damaging. No critical flaw jumped out in Theorems 1 and 2, and the numerical example is consistent with the theorems.\n\nMy recommendation: send it to peer review, but the referee should demand a rigorous converse for Theorem 3—either prove the product-form condition or explicitly characterize the region with the general auxiliaries and stop claiming tightness. As written, the central claim should not be accepted.","headline":"The inner and outer bounds are solid and the model is new, but the paper's headline tight characterization (Theorem 3) rests on a converse whose auxiliary variables do not satisfy the product-form distribution the theorem quantifies over.","tokens_in":33529,"tokens_out":3888,"would_cite":false,"duration_ms":37824,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["94A15","94A17","94A29"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves an exact secure strong-coordination region for conditionally independent sources and a deterministic legitimate channel in a multiple-access wiretap setting.","keywords":["strong coordination","multiple-access wiretap channel","secrecy","channel simulation","shared randomness","cribbing encoders","random binning","wiretap coding"],"falsifier":"Compute or simulate the secure strong-coordination region for a two-transmitter MAC-WT with a non-deterministic legitimate channel and sources with $I(X_1;X_2|W)>0$: if any rate pair outside Theorem 3's region is achievable, or any pair inside is not, the claimed completeness fails. More narrowly, for the paper's Example 1 without cribbing, an attempt to achieve secure channel simulation with $R_{01}<1$, for instance $R_{01}=0.75$, would disprove Proposition 1.","tokens_in":32449,"feed_emoji":"🔐","tokens_out":7119,"duration_ms":64227,"temperature":0.7,"pith_summary":"Two transmitters observe correlated sources and send over a multiple-access wiretap channel; a legitimate decoder, using shared randomness with each transmitter, must reproduce a target joint distribution with the sources while an eavesdropper learns nothing. The paper gives an achievable rate region and an outer bound on the shared-randomness rates needed for this secure strong-coordination task. Its main result is a complete characterization for the case where the two sources are conditionally independent given the decoder's side information and the legitimate channel consists of deterministic links: with one shared-randomness rate unlimited, the required rate is exactly $R_{01} \\ge I(U_1; X_1,Y | X_2,W,T) - H(\\tilde Y_1 | \\tilde Z,T)$, together with two entropy constraints on the channel outputs. If correct, this is a tight multi-terminal secure coordination result over a noisy wiretap channel for that setting, and it shows exactly how the eavesdropper's observation degrades the coordination rate.","feed_headline":"Exact secure-coordination rates found for wiretap MACs","feed_subtitle":"When sources are conditionally independent and the legitimate channel is deterministic, the paper pins down the shared-randomness rate.","key_machinery":"The argument is carried by two pairs of auxiliary random variables: $U_1$ and $U_2$ carry the source descriptions from Encoder 1 and Encoder 2 to the legitimate decoder, while $V_1$ and $V_2$ generate the wiretap code that hides the coordinated actions from the eavesdropper; $T$ is a time-sharing variable. Achievability is proved by random binning: the binning protocol is shown, via output-statistics-of-random-binning estimates and simultaneous decoding of correlated source descriptions, to induce almost the same joint distribution as a random-coding protocol, after which extra shared randomness is removed with a randomness-extraction lemma. The deterministic legitimate channel $\\tilde Y = (f_1(\\tilde X_1), f_2(\\tilde X_2))$ is what makes the characterization tight: setting $V_1 = \\tilde Y_1$ and $V_2 = \\tilde Y_2$ collapses the general inner bound to three constraints, and with conditional independence the matching converse is obtained directly. Cardinality bounds on $U_1$, $U_2$, and $T$ are established by a perturbation argument so that the converse region is compact and continuous.","core_discovery":"The paper's central discovery is Theorem 3: for conditionally independent sources ($I(X_1;X_2|W)=0$) and a legitimate channel of deterministic links $\\tilde Y = (f_1(\\tilde X_1), f_2(\\tilde X_2))$, the secure strong-coordination region with unlimited second shared-randomness rate $R_{02}$ is exactly the set of $R_{01}$ satisfying $H(\\tilde Y_1|T) \\ge I(U_1;X_1|W,T)$, $H(\\tilde Y_2|T) \\ge I(U_2;X_2|W,T)$, and $R_{01} \\ge I(U_1;X_1,Y|X_2,W,T) - H(\\tilde Y_1|\\tilde Z,T)$, for some auxiliary distribution whose marginal on $(X_1,X_2,W,Y)$ equals the target $q_{X_1X_2WY}$. The achievability side follows from the general inner bound by taking $V_1=\\tilde Y_1$, $V_2=\\tilde Y_2$, and a large $R_{02}$; the converse is proved directly rather than by specializing the general outer bound. This yields a matching inner-outer characterization for this multi-terminal noisy secure-coordination setting.","pith_inferences":["Beyond the paper, the structure of the tight result suggests that the general multiple-access wiretap coordination problem will resist single-letter solutions unless further independence or structural assumptions are imposed, since distributed source coding with decoder side information is tight only under such conditions.","The cribbing example indicates that one encoder's access to the other encoder's channel input can substitute for part of the shared randomness, so a cribbing link and a secret key may be traded off; the paper does not optimize this trade-off.","A natural testable extension is to vary the eavesdropper's noise level in the paper's example: the penalty term $H(\\tilde Y_1|\\tilde Z,T)$ predicts that a noisier wiretap observation monotonically reduces the shared-randomness rate required for secure coordination."],"forward_implications":["For conditionally independent sources and deterministic legitimate links, the secure coordination region is exactly known: no gap remains between achievable and converse rates when $R_{02}$ is unlimited.","The eavesdropper enters the single rate constraint only through the term $H(\\tilde Y_1|\\tilde Z,T)$, the residual entropy of the legitimate channel output given the wiretap observation; wiretap noise is thus priced explicitly in the shared-randomness rate.","Encoder cribbing can strictly enlarge the achievable region: in the paper's binary example, cribbing lowers the required channel entropy from $2$ to $1.5$ bits and the required shared-randomness rate $R_{01}$ from $1$ to $0.5$ bits per symbol.","The general inner and outer bounds (Theorems 1 and 2) still apply when the tight-case assumptions fail, so the paper supplies a fallback region for arbitrary correlated sources and noisy legitimate channels."],"supporting_citations":[{"why":"Establishes the strong-coordination framework and the converse-continuity techniques that the inner and outer bounds build on.","marker":"[7]"},{"why":"Supplies the output-statistics-of-random-binning achievability machinery used to prove the inner bound.","marker":"[29]"},{"why":"Provides the simultaneous-decoding bound for recovering the correlated source descriptions at the decoder.","marker":"[30]"},{"why":"Gives the single-user noisy-channel strong-coordination setting that this paper generalizes to multiple terminals with secrecy.","marker":"[15]"},{"why":"Provides the noiseless multiple-access coordination baseline whose tight independent-source result is extended to noisy wiretap channels.","marker":"[12]"},{"why":"Supplies the perturbation argument used to bound the cardinalities of the auxiliary random variables in the converse.","marker":"[35]"},{"why":"Defines the cribbing-encoder model used for the encoder-cooperation results in Section IV.","marker":"[4]"},{"why":"Supplies the total-variation lemmas and outer-bound tools for strong coordination over noisy channels with side information.","marker":"[16]"}],"fun_headline_variants":["Exact rates for secure coordination on wiretap MACs","Secure coordination: exact rate region for wiretap MACs","Tight secure-coordination rates found for wiretap links","Wiretap MAC coordination: exact rates achieved","Matching bounds for secure multi-terminal coordination"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The tight region collapses unless the sources are conditionally independent given the decoder side information and the legitimate channel outputs are deterministic functions of the two channel inputs; if either fails, only the non-matching inner and outer bounds remain.","fun_headline_variants_meta":{"raw":{"variants":["Exact rates for secure coordination on wiretap MACs","Secure coordination: exact rate region for wiretap MACs","Tight secure-coordination rates found for wiretap links","Wiretap MAC coordination: exact rates achieved","Matching bounds for secure multi-terminal coordination"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000233,"raw_usage":{"total_tokens":1582,"prompt_tokens":1122,"completion_tokens":460,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":738,"completion_tokens_details":{"reasoning_tokens":385}},"tokens_in":738,"tokens_out":460,"duration_ms":4024,"temperature":1.0,"reasoning_tokens":385,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:32:11.245669+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute or simulate the secure strong-coordination region for a two-transmitter MAC-WT with a non-deterministic legitimate channel and sources with $I(X_1;X_2|W)>0$: if any rate pair outside Theorem 3's region is achievable, or any pair inside is not, the claimed completeness fails. More narrowly, for the paper's Example 1 without cribbing, an attempt to achieve secure channel simulation with $R_{01}<1$, for instance $R_{01}=0.75$, would disprove Proposition 1.","supporting_citations":[{"cited_title":"The discrete memoryles s multiple-access channel with cribbing encoders,","cited_arxiv_id":null,"evidence_quote":"Defines the cribbing-encoder model used for the encoder-cooperation results in Section IV."},{"cited_title":"Distributed channel synthesis,","cited_arxiv_id":null,"evidence_quote":"Establishes the strong-coordination framework and the converse-continuity techniques that the inner and outer bounds build on."},{"cited_title":"Achievability proo f via output statistics of random binning,","cited_arxiv_id":null,"evidence_quote":"Supplies the output-statistics-of-random-binning achievability machinery used to prove the inner bound."},{"cited_title":"Simulation of a channel with another channel,","cited_arxiv_id":null,"evidence_quote":"Gives the single-user noisy-channel strong-coordination setting that this paper generalizes to multiple terminals with secrecy."},{"cited_title":"Multiple access channel simulation,","cited_arxiv_id":null,"evidence_quote":"Provides the noiseless multiple-access coordination baseline whose tight independent-source result is extended to noisy wiretap channels."},{"cited_title":"Evaluation of Marton’s inner bound for the general broadcast channel,","cited_arxiv_id":null,"evidence_quote":"Supplies the perturbation argument used to bound the cardinalities of the auxiliary random variables in the converse."},{"cited_title":"Stron g coordination of signals and actions over noisy channels wi th two-sided state information,","cited_arxiv_id":null,"evidence_quote":"Supplies the total-variation lemmas and outer-bound tools for strong coordination over noisy channels with side information."}],"review_version":1}