{"id":"a2d03ab1-0a82-4b91-b943-11bf615d2aaf","arxiv_id":"2411.19880","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"A low-power, drone-friendly QKD transmitter using RC-LEDs achieves state indistinguishability with eavesdropper mutual information around 1e-5 and a 532 kbit/s raw key rate.","lead":"The paper builds a compact quantum key distribution transmitter using three light-emitting diodes and measures how distinguishable the emitted quantum states are. It reports very low information leakage through the states' color and timing, plus a working lab link with raw key rates around 532 kilobits per second.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The reported indistinguishability values are lower-side estimates, not upper bounds: the mutual information is computed from histograms at finite spectral and temporal resolution, and the appendix explicitly assumes away unresolved fine structure that an eavesdropper could exploit.","rationale":"The reader's weakest-assumption identification is essentially correct: the indistinguishability estimate depends on the untested assumption of no unresolved fine structure, and the paper itself flags this in Appendix VIII A. I agree that this is the most load-bearing soft spot. I go slightly beyond the reader by noting that the spectral and temporal measurements are treated as separate marginals, so even fully within the measured resolution a joint spectral-temporal correlation could carry more information than either marginal alone; the appendix's fine-structure caveat covers this in spirit but not in the quantitative analysis. The reader's CONDITIONAL verdict already incorporates this concern, along with code-availability and secure-key-rate issues. My stress-test does not reveal a fatal internal inconsistency: the paper is an experimental characterization, not a claimed full security proof, and the limitation is openly acknowledged. The raw key rate is correctly labeled 'raw,' and the authors are careful not to claim a secure key rate. Given these considerations, the existing CONDITIONAL verdict is appropriate, and my read does not move it. The proposed joint high-resolution measurement would settle whether the acknowledged assumption actually holds, which is exactly the condition the paper's central claim rests on.","tokens_in":15403,"tokens_out":3467,"duration_ms":37316,"concrete_test":"Perform a joint high-resolution spectral-temporal characterization of the three RC-LED outputs at bright-light level, e.g., using a scanning Fabry-Perot interferometer with ~0.01 nm resolution together with a streak camera or frequency-resolved optical gating, and reconstruct the two-dimensional spectral-temporal probability distribution for each state. Then compute the joint mutual information I(B;E) from these 2D distributions using the same formalism as the appendix. If the joint value remains below about 1e-4 and no per-LED chirp or spectral fringes are observed, the assumption materially lands; if the joint value rises to ~1e-3 or higher, the reported leakage underestimates the side channel and the central claim would need to be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central, security-relevant claim is that the generated states are nearly indistinguishable, quantified by fractional mutual information values of 2.39e-5 (spectral) and 4.31e-5 (temporal). These values are computed from histograms recorded at 0.3 nm spectral resolution and roughly 170 ps timing jitter. Appendix VIII A states: 'we assume that thermal states describe the RC-LEDs without such identifying fine structure in the spectral-temporal modes,' and acknowledges that a faster detector, higher-resolution spectrometer, or more general pulse analysis could reveal distinguishing patterns. This makes the quoted figures conditional estimates, not security bounds. In a prepare-and-measure QKD side-channel analysis, the relevant quantity is an upper bound on information available to Eve in all non-polarization degrees of freedom; the present derivation provides a point estimate based on measured marginal distributions. A further, even within-resolution concern is that the spectral and temporal degrees of freedom are analyzed separately, so any joint spectral-temporal correlation (e.g., a per-LED chirp that is partially resolved but not captured by either marginal histogram) could make the joint mutual information larger than either value reported. The manuscript is transparent about the fine-structure assumption, and the physical argument that all LEDs share construction is reasonable; nevertheless, the headline 'highly indistinguishable states' as a security statement is only as strong as that assumption.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper presents a compact QKD transmitter that uses three resonant-cavity LEDs (RC-LEDs) as independent sources for a three-state BB84 polarization protocol with decoy states. The authors design and build a low-SWaP system using an FPGA for pulse generation, single-mode-fiber spatial filtering, a 1.2 nm spectral filter, and PLL-based temporal alignment to make the spectral and temporal degrees of freedom nearly indistinguishable. They quantify indistinguishability by computing the Shannon mutual information between an assumed eavesdropper and Bob from measured spectral and temporal histograms, obtaining values of 2.39e-5 (spectral) and 4.31e-5 (temporal), with bias corrections and uncertainties. In a tabletop QKD session at a 12.5 MHz clock rate, they report a raw key rate of 532 kbit/s and an average QBER of 1.83%. An appendix derives the mutual-information estimator and the bias/uncertainty treatment.","tokens_in":15673,"tokens_out":5027,"duration_ms":41899,"significance":"The paper addresses a real gap in low-SWaP QKD: multi-source prepare-and-measure systems often neglect or merely acknowledge side-channel distinguishability. The authors go further by quantifying leakage in spectral and temporal degrees of freedom, and by publishing the analysis code. The detailed treatment of measurement bias (Roulston correction) and Monte Carlo uncertainty for the spectral data is a strength. If the indistinguishability measurements are accepted as indicative rather than as security bounds, the system is a useful contribution to mobile QKD. However, the paper's central security-relevant claim requires that the reported values be upper bounds on leakage; the current analysis provides only conditional point estimates, and the reported rate is a raw key rate rather than a secure key rate.","major_comments":[{"comment":"The mutual information values are lower-side estimates, not upper bounds. The appendix states the assumption of no unresolved fine structure below measurement resolution and concedes that a faster detector or higher-resolution spectrometer could reveal distinguishing patterns. For a side-channel security claim, the relevant quantity is an upper bound on information available to Eve in all non-polarization DOFs. Please revise the abstract and conclusions to either explicitly present these figures as conditional estimates (e.g., \"under the assumption of no unresolved fine structure in the thermal RC-LED modes\") or add a robust upper-bounding argument. As written, the title and abstract overstate the security implication.","section":"Section VIII A; Section III B"},{"comment":"The spectral and temporal indistinguishabilities are quantified separately from marginal histograms. An eavesdropper could in principle exploit joint spectral-temporal correlations (e.g., a per-LED frequency chirp) that are not captured by either marginal distribution. The joint mutual information I(B; E_spectral, E_temporal | S) could exceed both reported values. Please either measure and report the joint distribution (or bound it), or provide a physical argument that no such correlations exist in the emitted modes.","section":"Section III B; Section VIII A"},{"comment":"The 532 kbit/s figure is a raw key rate, not a secure key rate. The paper does not report the sifted key rate, the error-correction throughput, the privacy-amplification cost, or a finite-key analysis with decoy-state parameter estimation. Since the stated design goal in Section II A is \"secure key rate,\" please clearly label the reported performance and, if possible, provide an estimated secure key rate for the tabletop configuration.","section":"Section IV A; abstract"},{"comment":"The expressions for p(B_k|S) contain leading minus signs. Since these are probabilities, the minus signs appear to be typographical errors that should be removed. If they are intentional, the derivation needs clarification. This is a presentation issue in a security-relevant derivation and should be fixed.","section":"Section VIII A, Eqs. (7) and (9)"}],"minor_comments":[{"comment":"The sentence beginning \"The probability of selecting the signal, non-zero decoy, and vacuum decoy states.\" is incomplete; it likely needs a verb or should be merged with the following sentence.","section":"Section III A 2"},{"comment":"\"PPL\" should be \"PLL\" in several places.","section":"Section III A 1; Section III B"},{"comment":"The phrase \"so that the intensity of decoy states is ∼40% lower than the signal states\" seems inconsistent with the stated currents: if the LED output is linear in current, 10 mA versus 25 mA gives 40% of the signal intensity (i.e., 60% lower), so please clarify the wording.","section":"Section III A 2"},{"comment":"The approximation switches from p_i to q_i in the denominator; defining q_i just before Eq. (20) or in the text immediately before Eq. (22) would improve readability.","section":"Section VIII A, Eq. (22)"},{"comment":"The caption refers to \"both signal and decoy states,\" but there are three states; please specify which states are shown.","section":"Figure 9b caption"},{"comment":"The term \"fractional mutual information\" is used, but the quantity in Eq. (4) is standard Shannon mutual information in bits (or nats); please clarify the units or the fraction being referenced.","section":"Abstract; Section VIII A"}],"recommendation":"major_revision","confidential_remarks":"The paper is within the journal's scope and the experimental effort is substantial, with commendable code availability. My main concern is the security framing: the indistinguishability figures are not upper bounds, and the reported key rate is raw. I would be willing to re-review a revised version that addresses these points."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The useful thing here is the measured leakage: fractional mutual information of 2.4×10^-5 (spectral) and 4.3×10^-5 (temporal) after filtering and timing compensation. That is a genuine step for the low-SWaP QKD subfield, where previous work acknowledged multi-source distinguishability but did not quantify it. The system integration is careful: three RC-LEDs, FPGA-driven pulses with decoy states, single-mode spatial filtering, Bayesian clock synchronization, and a tabletop QKD run with 1.83% QBER and 532 kbit/s raw rate. The numbers look internally consistent, and the appendix gives a clear derivation of the mutual information with bias corrections and uncertainties. Credit where due: the paper is transparent about its main assumption.\n\nThe soft spot is exactly what the appendix admits: the leakage values are computed from histograms at finite spectral and temporal resolution. A faster detector or higher-resolution spectrometer could in principle reveal per-LED fine structure, and the paper explicitly assumes it away. More subtly, the spectral and temporal degrees of freedom are analyzed separately, so a joint spectral-temporal correlation (say, a chirp that is partially resolved in neither marginal) could carry more information than either number alone. The quoted figures are therefore lower-side estimates, not upper bounds, and the headline \"highly indistinguishable\" should be read as \"indistinguishable under the stated assumptions.\" For a security-relevant side-channel claim, you would want an upper bound that includes unresolved structure. The physical argument that the LEDs share construction is reasonable, but it is not a proof. Also, the paper reports a raw key rate, not a secure key rate under a full security analysis, and reference [44] for code availability is a placeholder. These are all fixable.\n\nThis paper deserves a serious referee. I would send it back with three requests: state explicitly that the leakage values are not upper bounds unless the no-fine-structure assumption holds, and ideally give a bound that covers unresolved structure; report a secure key rate from a standard decoy-state analysis; and provide an actual code repository link. The work is a solid, honest engineering characterization, and it will be a useful citation for anyone building mobile QKD systems.","headline":"Useful systems paper that for the first time puts numbers on spectral and temporal side-channel leakage for a multi-source RC-LED QKD transmitter, but the numbers are conditional estimates, not security bounds.","tokens_in":16205,"tokens_out":2022,"would_cite":true,"duration_ms":19643,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["03.67.Dd"],"model":"deepseek-v4-flash","headline":"A compact RC-LED transmitter makes QKD states so indistinguishable that an eavesdropper gets only ~4e-5 bits of information per key bit.","keywords":["Quantum key distribution","BB84","resonant-cavity LED","state indistinguishability","side-channel attack","decoy state","free-space QKD","mobile platforms"],"falsifier":"Measure the spectral and temporal profiles of each RC-LED state with instruments of substantially higher resolution than those used here (e.g., an optical spectrum analyzer with <0.1-nm resolution and a detection chain with <10-ps jitter), recompute the mutual information from the resolved distributions, and check whether the per-state distinguishability rises above the reported $10^{-5}$ level.","tokens_in":15243,"feed_emoji":"🔐","tokens_out":5834,"duration_ms":47321,"temperature":0.7,"pith_summary":"This paper argues that a low size, weight, and power quantum key distribution transmitter built from resonant-cavity light-emitting diodes can make the quantum states nearly indistinguishable across the spatial, spectral, and temporal degrees of freedom, closing side channels that earlier mobile QKD designs left open. The authors quantify the residual information leakage to an eavesdropper as a fractional mutual information of $2.39\\times10^{-5}$ from spectral distinguishability and $4.31\\times10^{-5}$ from temporal distinguishability. They integrate the source into a full three-state BB84 decoy-state system and report a raw key rate of 532 kbit/s with average quantum bit error rate 1.83%. If correct, this shows that compact, commercially available LED sources and an FPGA-driven transmitter can support secure mobile QKD, such as between drones or automobiles, without sacrificing the state-indistinguishability needed for security.","feed_headline":"Mobile QKD source leaks just 4e-5 bits per key bit","feed_subtitle":"Drone-friendly transmitter hits 532 kbit/s raw key rate with 1.83% QBER and quantified side-channel leakage.","key_machinery":"The load-bearing elements are the three RC-LEDs (resonant-cavity light-emitting diodes) driven directly by FPGA-generated electrical pulses, combined with three filters on the unwanted degrees of freedom: a single-mode fiber that erases spatial distinguishability, a $1.2$-nm bandpass spectral filter that narrows the ~7-nm LED spectra, and sub-nanosecond (78-ps resolution) adjustable pulse timing that aligns the temporal wavepackets. The decoy-state protocol is used to secure the multi-photon thermal pulses that RC-LEDs emit. The indistinguishability is quantified by computing Shannon mutual information between Eve's measured spectral or temporal distributions and Bob's sifted bits, with a bias correction from entropy estimation theory.","core_discovery":"The central claim is that the three polarization states generated by the RC-LED-based transmitter can be made nearly indistinguishable in every degree of freedom except the intended one: the single-mode fiber makes the spatial modes identical, a $1.2$-nm spectral filter brings the spectra together, and FPGA-tuned electrical pulse delays and widths align the temporal waveforms. The measured fractional mutual information between an assumed eavesdropper and the legitimate receiver is $2.39\\times10^{-5}$ for spectral and $4.31\\times10^{-5}$ for temporal distinguishability, with the transverse spatial modes fully indistinguishable. The complete laboratory system achieves a raw key rate of 532 kbit/s and an average QBER of 1.83%, well below the 11% threshold for BB84-based protocols to yield a secure key.","pith_inferences":["If an eavesdropper could resolve per-LED spectral chirp or fine structure below the 0.3-nm and ~170-ps resolution, the true mutual information could exceed the reported values; a high-resolution heterodyne or ultrafast measurement of individual RC-LED pulses would settle this.","The same mutual-information quantification could be applied to other multi-source transmitters, such as VCSEL arrays or multiple lasers, to compare their side-channel resistance on a common scale.","Because the indistinguishability is engineered in the source, the leakage figures should transfer to field mobile links, although the raw key rate will drop with link loss and pointing errors.","The paper's assumption of thermal states without fine structure could be tested by measuring the second-order correlation function $g^{(2)}$ or the photon statistics of the filtered pulses, since any non-thermal structure would show up as a deviation."],"forward_implications":["A drone- or car-mounted QKD terminal can be built from commercially available RC-LEDs and an FPGA, with total transmitter size ~1800 cm³, weight ~2 kg, and power 2.5 W.","The quantified indistinguishability values give a concrete bound on side-channel leakage: an eavesdropper measuring only these degrees of freedom gains less than ~5e-5 bits of mutual information per sifted key bit.","The three-state BB84 protocol with decoy states is sufficient to turn the thermal, multi-photon RC-LED emission into a secure key at 532 kbit/s raw rate.","The QBER of 1.83% is far below the 11% threshold, so after error correction and privacy amplification a positive secure key rate is achievable.","The systematic measurement bias and uncertainty in the mutual information are small (below ~2.4e-5), so the leakage estimate is stable under the stated assumptions."],"supporting_citations":[{"why":"Supplies the decoy-state protocol that secures the multi-photon thermal emission of the RC-LEDs.","marker":"[24]"},{"why":"Establishes that a three-state BB84 protocol achieves the same secure key rate as four states, justifying the transmitter design.","marker":"[36]"},{"why":"Provides the loss-tolerant security analysis for QKD with imperfect sources, supporting the use of three states.","marker":"[37]"},{"why":"Introduces the resonant-cavity LED device that the transmitter is built around.","marker":"[32]"},{"why":"Shows that a single-mode fiber acts as a spatial filter, justifying the claim of fully indistinguishable transverse spatial modes.","marker":"[39]"},{"why":"Gives the entropy-estimation bias correction used to adjust the measured mutual information values.","marker":"[45]"},{"why":"Describes the photon-number-splitting attack that motivates the use of decoy states with imperfect sources.","marker":"[23]"},{"why":"Provides the Bayesian clock synchronization method used to align Alice's and Bob's data during sifting.","marker":"[40]"},{"why":"Represents a prior LED-based QKD system whose side-channel indistinguishability is not quantified, serving as the comparison this work addresses.","marker":"[12]"}],"fun_headline_variants":["Mobile QKD achieves 532 kbit/s, leakage under 5e-5","Drone-ready QKD: indistinguishable states, 1.83% QBER","QKD for drones: state leakage under 5e-5","Indistinguishable states enable secure drone QKD"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The leakage estimates assume the RC-LED wavepackets contain no identifying fine structure finer than the spectrometer's 0.3-nm resolution and the SPAD's ~170-ps jitter, so an eavesdropper with higher-resolution equipment could not see extra distinguishing features.","fun_headline_variants_meta":{"raw":{"variants":["Mobile QKD achieves 532 kbit/s, leakage under 5e-5","Drone-ready QKD: indistinguishable states, 1.83% QBER","QKD for drones: state leakage under 5e-5","Indistinguishable states enable secure drone QKD"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002041,"raw_usage":{"total_tokens":7958,"prompt_tokens":958,"completion_tokens":7000,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":574,"completion_tokens_details":{"reasoning_tokens":6922}},"tokens_in":574,"tokens_out":7000,"duration_ms":40443,"temperature":1.0,"reasoning_tokens":6922,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:42:20.304983+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the spectral and temporal profiles of each RC-LED state with instruments of substantially higher resolution than those used here (e.g., an optical spectrum analyzer with <0.1-nm resolution and a detection chain with <10-ps jitter), recompute the mutual information from the resolved distributions, and check whether the per-state distinguishability rises above the reported $10^{-5}$ level.","supporting_citations":[{"cited_title":"Experimental demonstration of drone-based quantum key distribution","cited_arxiv_id":"2302.14012","evidence_quote":"Supplies the decoy-state protocol that secures the multi-photon thermal emission of the RC-LEDs."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Establishes that a three-state BB84 protocol achieves the same secure key rate as four states, justifying the transmitter design."},{"cited_title":"Drone based superconducting single photon detection system with detection efficiency more than 90%","cited_arxiv_id":"2408.05878","evidence_quote":"Provides the loss-tolerant security analysis for QKD with imperfect sources, supporting the use of three states."},{"cited_title":"L¨ utkenhaus and M","cited_arxiv_id":null,"evidence_quote":"Introduces the resonant-cavity LED device that the transmitter is built around."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Shows that a single-mode fiber acts as a spatial filter, justifying the claim of fully indistinguishable transverse spatial modes."},{"cited_title":"(20) Thus, Eq","cited_arxiv_id":null,"evidence_quote":"Gives the entropy-estimation bias correction used to adjust the measured mutual information values."},{"cited_title":"Liu, X.-H","cited_arxiv_id":null,"evidence_quote":"Describes the photon-number-splitting attack that motivates the use of decoy states with imperfect sources."},{"cited_title":"Thus, it does not reveal any information to Eve because the basis and decoy state choices are already revealed over the public channel for the sifting stage of the analysis","cited_arxiv_id":null,"evidence_quote":"Provides the Bayesian clock synchronization method used to align Alice's and Bob's data during sifting."},{"cited_title":"06, 2024","cited_arxiv_id":null,"evidence_quote":"Represents a prior LED-based QKD system whose side-channel indistinguishability is not quantified, serving as the comparison this work addresses."}],"review_version":1}