Decoy state quantum key distribution with imperfect source

I. INTRODUCTION

• QKD implementations can expose backdoors when practical behavior deviates from security assumptions, allowing Eve to compromise theory-proved security.Such implementation vulnerabilities have been demonstrated across multiple QKD systems and may affect other quantum cryptography implementations using similar optical components.Source: §I · p. 1
• Weak coherent sources emit multiphoton pulses, enabling photon-number-splitting attacks that can reveal the final key after basis disclosure.Decoy-state protocols were introduced to counter this threat and have been implemented in many QKD systems.Source: §I · p. 1

II. DECOY STATE PROTOCOL

• The decoy-state protocol estimates signal-state single-photon detection and error behavior from gains and error rates of signal, weak-decoy, and vacuum intensities.For phase-randomized weak coherent sources, photon-number probabilities and linear constraints provide bounds on the single-photon yield and error rate used in the key-rate calculation.Source: §I · p. 1, §II · p. 2

I. INTRODUCTION

• The protocol assumes Eve cannot distinguish signal and decoy states beyond their mean photon numbers, but source imperfections can violate this assumption through side channels.Trace distance characterizes leaked information, and the resulting security analysis bounds the key rate while estimating single-photon yield and error rate under distinguishability.Source: §I · pp. 1–2
• This work tests pump-current and external intensity modulation, models a PNS attack exploiting timing differences, and derives analytical security estimates for distinguishable states.It also proposes calibrating Bob’s transmittance and applies the method to three QKD systems across transmission distance.Source: §I · p. 2

III. INTENSITY MODULATION TEST

• The pump-current and external-modulation methods generate signal and decoy states by modulating the laser diode directly or with a downstream intensity modulator, respectively.Both implementations use fixed attenuation to reach the single-photon level.Source: §III · p. 3
• Pump-current modulation produces mismatched signal and decoy timing distributions, whereas external modulation shows no measurable timing mismatch.The signal emits earlier and has a secondary peak from 662 to 937 ps, while the decoy distribution drops in that interval.Source: §III · pp. 3–4, Figure 2 · p. 4
• The pump-current mismatch arises because higher current triggers earlier laser emission and additional relaxation-oscillation pulses before the current is interrupted.The lower-current decoy pulse starts later and produces only one light pulse.Source: §III · pp. 3–4, Figure 1 · p. 3

IV. PNS ATTACK

• Eve exploits partial distinguishability by selecting time windows Ws and Wd, classifying pulses, and selectively blocking, forwarding, or storing photons during a PNS attack.The attack manipulates yields while using measured guessing probabilities and optimized windows to preserve detection statistics.Source: §IV · pp. 4–6
• Beyond 45 km, the optimized upper key-rate bound falls below the protocol’s lower bound, allowing Eve to eavesdrop part of the secret key.The simulation uses GYS detection parameters and optimizes Ws and Wd separately for each distance.Source: §IV · pp. 4, 6, Figure 3 · p. 6

V. SECURE KEY RATE WITH IMPERFECT SOURCE

• The paper generalizes the decoy-state model to source imperfections in time, frequency, phase, or other dimensions that may let Eve distinguish signal from decoy pulses.It derives two analytical formulas for estimating single-photon yield and error-rate contributions under such distinguishability.Source: §V · p. 6

A. Model

• The weak-plus-vacuum protocol uses intensities ω = {µ, ν, ν1}, with µ > ν > ν1 and µ > ν + ν1.ν1 represents the vacuum-state intensity.Source: A. Model · p. 6
• Source imperfections are modeled by an Eve-accessible parameter λ and intensity-dependent states ρω(λ) distributed according to normalized functions fω(λ).λ can represent time, frequency, or another measured distinguishing dimension.Source: A. Model · pp. 6–7
• The model assumes λ is independent of photon number, so conditional yields Yn(λ) and error rates en(λ) depend on λ but not directly on the pulse intensity.When ρµ(λ) = ρν(λ) = ρν1(λ), the standard decoy-state method is recovered.Source: A. Model · p. 7
• The source imperfection is quantified by the trace distance between the intensity-conditioned states ρω(λ) and ρω′(λ).Trace distance is used to characterize how distinguishable the source states are.Source: A. Model · p. 7

B. Lower bound of Y µ

• The derivation follows the standard weak-plus-vacuum decoy procedure and assumes equal vacuum yields Yµ_0 = Yν_0 = Yν1_0 = Y0.The assumption reflects that Eve obtains no information from vacuum pulses.Source: B. Lower bound of Y µ · p. 7
• The imperfect-source lower bound is written as G(µ, ν, ν1) − g(µ, ν, ν1), where G is the standard bound and g represents information leaked by source imperfection.The standard term G is unchanged from the perfect-source lower bound.Source: B. Lower bound of Y µ · p. 7
• For the weak-plus-vacuum case, ν1 = 0, and the derivation notes that removing a factor of 2 could improve the key rate.The bound is obtained from inequalities over the intensity-dependent source states.Source: B. Lower bound of Y µ · p. 7

• The upper bound of e^µ_1 is derived analytically for distinguishable signal and decoy states.Source: C. Upper bound of eµ · p. 7
• When D_µν1 = 0, Eq. (24) reduces to the standard decoy-state upper bound.Source: C. Upper bound of eµ · p. 7
• For the weak + vacuum protocol, ν1 = 0 and Y0 = Qvac simplify the upper-bound expression for e^µ_1.Source: C. Upper bound of eµ · p. 7

D. Numerical simulation

• The imperfect-source simulation estimates single-photon yield, error rate, and key rate while optimizing signal and decoy intensities over specified ranges.Only cases with positive final key rate are shown; D_µν = 10^-1 yields no secure key.Source: Figure 4 · p. 8
• The maximum secure distance falls from about 141 km for D_µν = 0 to 124, 92, and 48 km for D_µν = 10^-5, 10^-4, and 10^-3.No positive key rate is possible for D_µν = 10^-2 or 10^-1.Source: D. Numerical simulation · pp. 7–8, Figure 4 · p. 8

E. Theory improvement

• Calibrating Bob’s optical transmittance improves the final key rate and maximum distance under the advanced security proof.The calibration assumption is motivated by Bob’s protected receiver devices and active transmittance calibration.Source: E. Theory improvement · p. 8
• For D_µν = 10^-3, the improved proof increases maximum distance from 48 km to 105 km.It also provides positive key rates up to 64 km for D_µν = 10^-2 and 18 km for D_µν = 10^-1.Source: Figure 5 · p. 8, E. Theory improvement · p. 9
• The calibrated-transmission assumption does not apply to MDI QKD because its detection part is outside the protected zone and can be controlled by Eve.Source: E. Theory improvement · p. 9

VI. DISCUSSION AND APPLICATION EXAMPLES

• Dµν = 0.4005 produces zero secure key rate under both security proofs, showing that sufficiently large modulation imperfection can make the system insecure.For the improved proof, the receiver transmittance is assumed to be ηBob = 4.5%.Source: §VI · p. 9
• For Dµν = 3.6 × 10^-3, the maximum distance is 22 km under the initial proof and 83 km under the advanced proof.The advanced-proof result assumes ηBob = 4.5%; positive key generation is possible despite the low mismatch, but distance remains sensitive to mismatch.Source: §VI · p. 10
• The analysis can accommodate distinguishability in arbitrary dimensions, but the experiment measured only time-domain mismatch and did not measure possible spectral mismatch.Gain-switched lasers can exhibit chirp, and both modulation methods may introduce joint time–frequency distinguishability.Source: §VI · pp. 9–10
• For Dµν = 0.1400, the initial proof yields no positive key rate, whereas the improved proof supports secure transmission up to 10 km.The joint time–frequency distribution is estimated by assuming independence because time-resolved spectroscopy was not performed.Source: Figure 6 · p. 9, §VI · p. 10
• The external intensity modulator has the smallest signal–decoy mismatch, yielding the highest key rate and longest transmission distance among the three implementations.The comparison assumes the different sources operate in the same fibre-based QKD system with identical detection-side parameters.Source: §VI · pp. 9–10

VII. CONCLUSION

• The method provides analytical estimates for single-photon yield and error rate, but its current scope excludes distinguishability among BB84 states and separately derived security properties for MDI QKD.Measured mismatch values are treated as real mismatch, making the reported key rates conservative estimates.Source: §VI · pp. 9–10, §VII · p. 11

• The appendix first estimates a lower bound for the background rate Y0, using the vacuum-decoy relation when applicable.Source: Appendix A · pp. 11–12
• A lower bound for Y1 is then obtained by solving the stated inequality under the assumption µ ≥ ν + ν1.Source: Appendix A · p. 12

• The appendix derives an upper bound for the single-photon error rate eµ_1 from the defining relation for e1 and the distinguishable-decoy bounds.Source: Appendix B · pp. 12–13
• The resulting expression provides the upper bound used to estimate eµ_1 for the decoy-state protocol.Source: Appendix B · pp. 12–13
• Vacuum pulses are treated identically across intensities, so Yµ_0 = Yν_0 and eµ_0 = eν_0; for weak-plus-vacuum decoys, ν1 = 0 and Y0 = Qvac.Source: Appendix B · pp. 12–13