{"id":"9e0fc3a3-4941-41a7-8723-1618077616b4","arxiv_id":"2506.19438","paper_version":3,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":1,"one_line_summary":"A practical squeezed-state CV-QKD system over optical fiber, using local-local oscillator and DSP, achieves higher finite-size secret key rates than coherent-state CV-QKD under low reconciliation efficiency and high excess noise.","lead":"This paper demonstrates a fiber-based quantum key distribution system that uses squeezed light instead of regular laser light, and shows it can generate secret keys faster and tolerate more noise than standard approaches. The result suggests squeezed states, long considered too fragile for practical use, are now a viable option for secure fiber communication.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The squeezed-state advantage rests on treating ~3 SNU of anti-squeezed preparation noise as trusted; without quantifying the untrusted case or the stability of the B2B noise-floor calibration, the finite-size key rates in Figs. 2–3 may be optimistic.","rationale":"The experiment itself is credible: real fiber transmission, LLO with DSP-based phase recovery, finite-size analysis, and a fair single-quadrature-versus-two-quadrature comparison are all genuine advances. However, the quantitative headline — higher key fractions and greater noise tolerance than coherent states — depends on treating a large preparation impurity as trusted noise. The paper explicitly flags the alternative ('impurity can be incorporated into the security analysis' but 'reduces performance') and points to the untrusted-noise analysis in Ref. [27], yet it does not apply that analysis to the demonstrated parameters. Since the coherent-state benchmark has no analogous trusted-noise subtraction, the comparison is not fully symmetric unless the trusted-noise assumption is justified by continuous monitoring data. The B2B estimation procedure is a second, related vulnerability: any drift between calibration and key extraction directly biases the excess-noise estimate, and the finite-size confidence intervals in Eqs. (3)–(6) do not cover systematic calibration errors. The proposed test directly quantifies whether the advantage survives without the trusted-noise assumption. The reader's weakest assumption identified the same point; the verdict remains CONDITIONAL because the concern is real but does not invalidate the experimental achievement or the protocol concept.","tokens_in":16818,"tokens_out":7577,"duration_ms":81303,"concrete_test":"Recompute the finite-size secret key fraction for the Table I 50-km parameters (Vsqz = 0.417 SNU, ΔVAN = 3.029 SNU, VM = 1.372 SNU, η = 0.166, τ = 0.68, ε ≈ 0.04 SNU, n = 10^8) using the untrusted anti-squeezed-noise analysis referenced in Supplementary Material S.I (Ref. [27]/[S2]), i.e., purify ΔVAN through Eve rather than Alice. If the squeezed-state key fraction drops to or below the coherent-state value at any β in Fig. 2(a), the reported 6x advantage is contingent on the trusted-noise assumption. If that model is not readily available, an equivalent check is to repeat the key-rate calculation with ΔVAN treated as untrusted in the EB model of Fig. S1, and to propagate a ±10% drift in the B2B-measured floor through Eq. (S8) to see whether the 6.5σ excess-noise interval remains positive.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Table I reports ΔVAN = 3.029 SNU, which is larger than the pure anti-squeezed variance 1/Vsqz ≈ 2.398 SNU. The three-squeezer EB purification in Fig. S1 (from Ref. [27]) and the key-rate Eq. (6) are valid only if this noise is trusted, i.e., its purification is held by Alice and not by Eve. The paper's only support is a B2B characterization (Supp. S.II); Sec. III.B states that 'the noise floor without induced excess noise is estimated rather than directly measured.' There is no demonstrated continuous monitoring of Vsqz or ΔVAN during the 10^8-symbol run, and Supp. S.II itself warns that the channel is under Eve's control, so the floor cannot be measured through the channel. Because the estimated excess noise is tiny (εx, εp ≈ 0.04 SNU at channel input), a small bias in the subtracted floor (V_Rx_sqz and V_Rx_anti + τη/2 ΔVAN in Eq. S8) translates into a large relative error in ε_up and hence in K_finite. The text acknowledges that untrusted impurity 'reduces performance' (Sec. II.B) but never quantifies that reduction for the demonstrated parameters. The coherent-state comparison does not involve subtracting a 3-SNU trusted floor, so the asymmetry of the comparison is exactly where the central claim is least secure.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper reports an experimental demonstration of a continuous-variable quantum key distribution (CV-QKD) system using quadrature-squeezed states over optical fiber, with a local local-oscillator (LLO) and a digital-signal-processing (DSP) chain for carrier recovery. Alice prepares Gaussian-modulated squeezed vacuum states, Bob performs RF heterodyne detection and publicly announces the anti-squeezed quadrature, and the squeezed quadrature is used for key generation. The authors estimate channel parameters from about 10^8 quantum symbols, compute finite-size secret key fractions against collective attacks using a security framework from their own earlier work (Refs. [26,27]), and compare the squeezed-state protocol with a coherent-state protocol under the same conditions. They report higher key fractions at lower reconciliation efficiencies and a higher tolerance to excess noise for the squeezed-state protocol, and conclude that this is the first practical fiber-based squeezed-state CV-QKD demonstration.","tokens_in":17120,"tokens_out":4297,"duration_ms":40405,"significance":"If the security analysis withstands scrutiny, this is a significant experimental step: it moves squeezed-state CV-QKD from tabletop proof-of-principle setups to a fiber-compatible, LLO-based implementation, and it provides a quantitative finite-size comparison with coherent states. The detailed parameter tables, the reconciliation-efficiency study, and the excess-noise tolerance comparison are valuable contributions. The main caveat is that the claimed advantage depends on modeling the large anti-squeezed preparation noise as trusted noise under Alice's control, which is not yet justified by a demonstrated continuous-monitoring procedure. This issue is load-bearing for the central claim and needs to be addressed explicitly.","major_comments":[{"comment":"The central advantage claim relies on treating the anti-squeezed preparation noise ΔV_AN = 3.029 SNU (Table I) as trusted noise under Alice's control, via the three-squeezer purification of Fig. S1. The text states in Sec. II.B that continuous monitoring of squeezing and purity would justify this treatment, but no such monitoring is demonstrated during the 10^8-symbol run, and Supp. S.II only describes a back-to-back (B2B) calibration. Because the estimated channel excess noise is only ~0.04 SNU, even a small error in attributing ΔV_AN would substantially change the upper bound on excess noise and hence the finite-size key rates in Figs. 2 and 3. Please quantify the reduction in key rate if ΔV_AN is instead treated as untrusted noise (with the purification held by Eve), or provide a monitoring protocol with proven security that justifies the trusted-noise assumption.","section":"Sec. II.B and Table I"},{"comment":"The noise floor for the squeezed-state protocol is estimated, not directly measured: Eq. (S8) subtracts V^Rx_sqz and V^Rx_anti-sqz + (τη/2)ΔV_AN, where V^Rx_sqz and V^Rx_anti-sqz are derived from the B2B calibration and the estimated channel efficiency. As Supp. S.II concedes, the channel is under Eve's control, so this floor cannot be validated through the channel. Any drift in the squeezing source or in detection efficiency between the B2B measurement and the transmission would bias the excess-noise estimates; given the small excess-noise values (~0.04 SNU at channel input), the resulting relative error in the finite-size key rate could be large. The authors should provide a stability analysis of the B2B calibration (e.g., repeated measurements over time) or a sensitivity analysis showing how much key-rate error would result from a specified drift.","section":"Supp. S.II and Sec. III.B"},{"comment":"The theory curves in Fig. 3 are described as a 'fitted theory model,' but the fitting parameters and procedure are not specified. It is therefore unclear whether the agreement between the data points and the curves is a genuine test of the model or a consequence of adjustable parameters. Please state which quantities were fitted, their fitted values, and the residuals, or else present the comparison as parameter-free predictions based on the independently estimated parameters.","section":"Sec. III.B, Fig. 3"}],"minor_comments":[{"comment":"In Eq. (4), 'quaratures' should be 'quadratures'.","section":"Eq. (4)"},{"comment":"In Sec. III.A, 'with an total loss of of 7.25 dB' should read 'with a total loss of 7.25 dB' (duplicate 'of').","section":"Sec. III.A"},{"comment":"In Sec. III.A, 'at β≈0.89%' should be 'at β≈0.89' (without the percent sign).","section":"Sec. III.A"},{"comment":"The captions for Tables S3 and S4 state 'estimated from 108 quantum symbols'; please correct this to 10^8 quantum symbols.","section":"Supp. Tables S3 and S4"},{"comment":"In Table S3, the entry 'blue0' in the SKF column for the coherent-state row at excess noise 0.084 appears to be a formatting artifact; please replace it with the numerical value or a dash.","section":"Supp. Table S3"},{"comment":"In Sec. II.B, 'Details analysis' should be 'Detailed analysis'.","section":"Sec. II.B"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a strong experimental advance, but the security claim is only as solid as the trusted-noise assumption. I recommend insisting on a quantitative untrusted-noise analysis or a demonstrated monitoring scheme before publication. The authors' reliance on their own prior framework (Refs. [26,27]) is not circular, but independent validation or a worst-case analysis would considerably strengthen the paper."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You should know this paper is the first real fiber-based demonstration of squeezed-state CV-QKD with a locally generated local oscillator and DSP-based phase recovery. Prior squeezed-state experiments used free-space links or emulated loss, so this is a genuine step forward. The authors build a 1550-nm squeezed source, modulate with an IQ modulator, send over 50 km of ULL fiber, and recover phase using an unscented Kalman filter plus squeezing-angle estimation and quadrature remapping. They compare against coherent states under identical conditions by simply blocking the squeezed light. That comparison is well done and underpins the paper's main claims: better key rate at realistic reconciliation efficiency and better tolerance to excess noise. The parameter tables are detailed and the finite-size analysis follows their published framework.\n\nWhere I get uneasy is the security analysis, and it is the same place the authors are least explicit. Table I lists ΔVAN = 3.029 SNU, which is larger than the pure anti-squeezed variance 1/Vsqz ≈ 2.398 SNU. The key-rate formulas treat all of that as trusted noise under Alice's control, with the purification held by Alice. If Eve holds any of it, the finite-size key rates are overestimated. The paper says untrusted impurity \"reduces performance\" and cites their own Ref. [27] for the untrusted analysis, but it never quantifies the reduction at the demonstrated parameters. Since the excess noise numbers are tiny (≈0.04 SNU), a small bias in the back-to-back noise-floor subtraction could matter. They are transparent that the floor is \"estimated rather than directly measured\"; that honesty is welcome, but it does not close the gap.\n\nAlso, the theory curves in Fig. 3 are described as fitted, there are no error bars on the key-rate points, and no data or code are released. These are minor-to-moderate issues on top of the trusted-noise point.\n\nNone of this makes me think the central result is wrong. The qualitative advantage of squeezed states under low reconciliation efficiency and high excess noise is consistent with earlier theory and with the experimental curves. But the quantitative advantage—six to twenty times—should be read with caution until the untrusted case is quantified or the trusted-device assumptions are made explicit.\n\nThis paper deserves a serious referee. I would send it to review and ask specifically for (i) a plot or table of the key rate with ΔVAN treated as untrusted, using the measured parameters, and (ii) a stability analysis of the B2B noise floor over the full run. For anyone working in CV-QKD, this is worth reading and discussing.","headline":"First genuine fiber-based squeezed-state CV-QKD demonstration with LLO/DSP, but the quantitative advantage over coherent states rests on treating ~3 SNU of anti-squeezed preparation noise as trusted, which the paper never quantifies in the untrusted case.","tokens_in":17669,"tokens_out":3159,"would_cite":true,"duration_ms":31625,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper reports the first practical fiber-based continuous-variable quantum key distribution system using quadrature-squeezed states, and shows it outperforms coherent-state systems in finite-size secret key rate and excess-noise…","keywords":["continuous-variable quantum key distribution","squeezed states","finite-size security","local local oscillator","digital signal processing","heterodyne detection","excess noise","reconciliation efficiency"],"falsifier":"Take the same 50 km fiber setup but, instead of estimating the noise floor from a back-to-back measurement, verify it with an independent direct measurement while Eve is free to vary channel loss, and re-compute the finite-size key rate; if the key fraction drops to zero under the corrected noise floor, the claimed advantage over coherent states fails.","tokens_in":16638,"feed_emoji":"🔐","tokens_out":5748,"duration_ms":51784,"temperature":0.7,"pith_summary":"Quantum key distribution with continuous variables has practical appeal because it uses standard telecom equipment, but most systems encode information in coherent states. This paper reports the first practical fiber-based implementation that instead uses quadrature-squeezed states, with a locally generated local oscillator and a digital-signal-processing chain that removes the need for optical phase locking. Over a 50 km fiber link, the squeezed-state protocol achieves finite-size secret key fractions several times higher than a coherent-state system under identical conditions, and it keeps producing keys at reconciliation efficiencies and excess-noise levels where the coherent-state system stops. A sympathetic reader would take away that squeezed states have moved from a theoretical advantage to a working resource for real QKD deployments.","feed_headline":"Squeezed states beat coherent states in first fiber QKD test","feed_subtitle":"A squeezed-state CV-QKD system over 50 km yields up to 20x higher key throughput.","key_machinery":"The central object is the Gaussian squeezed-state protocol with heterodyne detection and public announcement of the anti-squeezed quadrature. In this protocol, Alice and Bob keep only the squeezed quadrature for key generation, while the anti-squeezed quadrature is used for parameter estimation; this is captured in the three-mode entanglement-based purification model, with the Holevo bound conditioned on the announced anti-squeezed measurement. Two additions carry the work: a DSP chain that uses a frequency-multiplexed pilot tone and an unscented Kalman filter to recover the carrier phase without optical locking, and a parameter-estimation extension that explicitly includes detector efficiency, giving the estimators in the paper's Eqs. (3)-(5) and the finite-size key fraction in Eq. (6).","core_discovery":"The central claim is that quadrature-squeezed states can serve as the signal states in a practical, fiber-based CV-QKD protocol, and that they outperform coherent states when both are analysed with finite-size security against collective attacks over the same channel. The authors demonstrate this with a prepare-and-measure setup in which Alice generates squeezed vacuum at 1550 nm, displaces it with Gaussian modulation in both quadratures, and sends it over 50 km of ultra-low-loss fiber; Bob uses a free-running laser as local oscillator and RF heterodyne detection, and a DSP chain recovers the carrier phase and re-aligns the measured quadratures. Only the squeezed quadrature is used for the key; the anti-squeezed quadrature is publicly announced for parameter estimation. With all parameters estimated from $10^{8}$ quantum symbols, the squeezed-state protocol yields up to 0.0105 bits/symbol, roughly six times the coherent-state rate at high reconciliation efficiency, and it remains positive at efficiencies and excess noises where the coherent-state protocol yields zero key.","pith_inferences":["If the trusted-noise assumption for the anti-squeezed quadrature is accepted, the security analysis could be extended to discrete-modulated squeezed states; the paper notes the theory for this is lacking, but the experimental toolkit here is ready.","The reported advantage depends on using the squeezed quadrature alone for keys; a system that tries to use both quadratures would likely lose much of the benefit, since heterodyne detection adds trusted vacuum noise that effectively counteracts channel noise.","A direct measurement of the channel noise floor (rather than a back-to-back estimate) could test how much of the claimed margin survives realistic channel drift, especially over long deployment times."],"forward_implications":["Fiber-based CV-QKD can now be built with squeezed states as a modular replacement: the squeezed-light source is added to a coherent-state transmitter, and the receiver needs only a free-running local oscillator and DSP.","Squeezed-state systems can maintain positive key rates with lower reconciliation efficiencies, reducing the computational load of error correction; the reported GPU-based reconciliation yields 30.19 kbit/s for squeezed states versus 1.57 kbit/s for coherent states.","Squeezed states tolerate higher excess noise, which suggests they can operate alongside high-power classical channels in a shared fiber, a regime where coherent-state CV-QKD fails.","The same DSP-based approach is directly applicable to free-space and satellite channels, where the paper notes squeezed states are predicted to be even more advantageous."],"supporting_citations":[{"why":"Provides the theoretical prediction that squeezed states improve key rates and noise tolerance, which this paper sets out to verify experimentally.","marker":"[26]"},{"why":"Supplies the finite-size security analysis and the three-mode purification model for the squeezed-state protocol, including the trusted anti-squeezed noise treatment.","marker":"[27]"},{"why":"The compact squeezed-light source used on the transmitter side.","marker":"[28]"},{"why":"The unscented Kalman filter method for carrier phase recovery that the DSP chain is built on.","marker":"[38]"},{"why":"Demonstrates digital reconstruction of squeezed states and passive CV-QKD, the basis for the receiver DSP.","marker":"[41]"},{"why":"Quadrature remapping technique used to align Alice's and Bob's modulation bases.","marker":"[45]"},{"why":"The MET-LDPC error correction code with puncturing used for information reconciliation at varying efficiencies.","marker":"[49]"}],"fun_headline_variants":["Squeezed states boost fiber QKD key rates 20x over coherent","Squeezed-light CV-QKD outperforms coherent over 50 km fiber","Practical squeezed QKD: higher key rate, resilience to noise","Squeezed QKD achieves 20x key throughput in fiber test"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole scheme relies on Alice's anti-squeezed preparation noise being trusted and faithfully monitored; if an eavesdropper can influence or mimic that noise, the finite-size key rates claimed here would not hold.","fun_headline_variants_meta":{"raw":{"variants":["Squeezed states boost fiber QKD key rates 20x over coherent","Squeezed-light CV-QKD outperforms coherent over 50 km fiber","Practical squeezed QKD: higher key rate, resilience to noise","Squeezed QKD achieves 20x key throughput in fiber test"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000266,"raw_usage":{"total_tokens":1581,"prompt_tokens":888,"completion_tokens":693,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":614}},"tokens_in":504,"tokens_out":693,"duration_ms":6858,"temperature":1.0,"reasoning_tokens":614,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T18:32:45.650579+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take the same 50 km fiber setup but, instead of estimating the noise floor from a back-to-back measurement, verify it with an independent direct measurement while Eve is free to vary channel loss, and re-compute the finite-size key rate; if the key fraction drops to zero under the corrected noise floor, the claimed advantage over coherent states fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the theoretical prediction that squeezed states improve key rates and noise tolerance, which this paper sets out to verify experimentally."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the finite-size security analysis and the three-mode purification model for the squeezed-state protocol, including the trusted anti-squeezed noise treatment."},{"cited_title":"Arnbak, C","cited_arxiv_id":null,"evidence_quote":"The compact squeezed-light source used on the transmitter side."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The unscented Kalman filter method for carrier phase recovery that the DSP chain is built on."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Demonstrates digital reconstruction of squeezed states and passive CV-QKD, the basis for the receiver DSP."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Quadrature remapping technique used to align Alice's and Bob's modulation bases."},{"cited_title":"Quantum Security Networks Partnership","cited_arxiv_id":null,"evidence_quote":"The MET-LDPC error correction code with puncturing used for information reconciliation at varying efficiencies."}],"review_version":2}