Photon-Number-Splitting (PNS) Attack
The Photon-Number-Splitting (PNS) attack targets quantum communication systems that use multi-photon pulses. In this strategy, a threat actor (Eve) exploits the fact that these pulses can contain more than one photon carrying the same information, allowing them to split off one photon, measure it, and leave the communication seemingly undisturbed.
Because the threat actor (Eve) also holds back just enough of the traffic to leave the receiver (Bob) counting exactly the number of arrivals the lossy link would have delivered, nothing in the error rate or the detection rate reveals her. The result is not a broken link but a silently readable one: where the source emits enough multi-photon pulses, she ends the session holding the key while the transmitter (Alice) and the receiver (Bob) see a normal run [Brassard2000,Liu2011].
| [Ashkenazy2024] | Ariel Ashkenazy, Yuval Idan, Dor Korn, Dror Fixler, Barak Dayan, Eliahu Cohen. "Photon Number Splitting Attack - Proposal and Analysis of an Experimental Scheme" (2024) 10.1002/qute.202300437. https://onlinelibrary.wiley.com/doi/abs/10.1002/qute.202300437. |
| [Avanesov2025] | Avanesov, Kronberg. "Generalized photon-number-splitting attack on a coherent one-way quantum-key-distribution protocol" (2025) 10.1103/PhysRevA.111.042601. https://doi.org/10.1103/PhysRevA.111.042601. Created at abstract depth. The full text is behind the APS paywall, has no arXiv version, and is not in the local library; only the abstract was read, from the publisher's landing page. Authors are recorded by surname only and the abstract text is not stored. Because no quote from this paper can be verified against a primary source, the A-0044 review deliberately names it in prose rather than with an inline bracket marker; add the markers once the PDF is obtained, along with the full author names.. |
| [Bennet1992] | C.H. Bennett, F. Bessette, G. Brassard, L. Salvail, J. Smolin. "Experimental quantum cryptography" In: J. Cryptology 5, 3–28. (1992) 10.1007/BF00191318. |
| [Bennett1984] | C. H. Bennett, G. Brassard. "Quantum cryptography: Public key distribution and coin tossing" In: Proceedings of IEEE International Conference on Computers, Systems and Signal Processing, volume 175, page 8. New York, 1984. (1984) 10.1016/j.tcs.2014.05.025. |
| [Brassard2000] | G. Brassard, N. Lutkenhaus, T. Mor, and B. Sanders. "Limitations on Practical Quantum Cryptography" In: Phys. Rev. Lett. 85 (2000), p. 1330. (2000) 10.1103/PhysRevLett.85.1330. |
| [BSI2023] | BSI. "Implementation Attacks against QKD Systems". (2023) https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Publications/Studies/QKD-Systems/QKD-Systems.pdf. |
| [CurrasLorenzo2022] | G. Curras-Lorenzo, K. Tamaki, and M. Curty. "Security of quantum key distribution with imperfect phase randomisation" In: Quantum Sci. Technol. 9, 015025. (2024) 10.1088/2058-9565/ad141c. arXiv:2210.08183. |
| [Felix2001] | S. Felix, N. Gisin, A. Stefanov, and H. Zbinden. "Faint laser quantum key distribution: eavesdropping exploiting multiphoton pulses" In: Journal of Modern Optics 48 (2001), pp. 2009–2021. (2001) 10.1080/09500340108240903. |
| [Gabdulkhakov2018] | I. Gabdulkhakov, O. Morozov, G. Morozov, M. Zastela, A. Tyajelova, and L. Sarvarova. "Frequency coding quantum key distribution channel based on serial photons amplitude modulation and phase commutation" In: Proc. SPIE 10774, Optical Technologies in Telecommunications 2017, 107741Q. (2018) 10.1117/12.2322488. |
| [Gottesman2004] | D. Gottesman, H.-K. Lo, N. Lütkenhaus, J. Preskill. "Security of quantum key distribution with imperfect devices". In: Quantum Information and Computation Vol.4 No.5. (2004) arXiv:quant-ph/0212066. |
| [Krelina2025] | Michal Krelina, Tom Sorger, Bob Dirks. "SQOUT: A Risk-Based Threat Analysis Framework for Quantum Communication Systems" (2025) 10.1016/j.cose.2026.105034. arXiv:2510.23462. https://doi.org/10.1016/j.cose.2026.105034. |
| [Liu2011] | W. Liu, S. Sun, L. Liang, and J. Yuan. "Proof-of-principle experiment of a modifed photon-number-splitting attack against quantum key distribution" In: Phys. Rev. A 83 (2011), p. 042326. (2011) 10.1103/PhysRevA.83.042326. |
| [Lutkenhaus2000] | N. Lutkenhaus. "Security against individual attacks for realistic quantum key distribution" In: Phys. Rev. A 61 (2000), p. 052304. (2000) 10.1103/PhysRevA.61.052304. |
| [Lutkenhaus2002] | N. Lutkenhaus and M. Jahma. "Quantum key distribution with real- istic states: photon-number statistics in the photon-number splitting attack" In: New Journal of Physics 4 (2002), p. 44. (2002) 10.1088/1367-2630/4/1/344. |
| [Mailloux2016] | L. O. Mailloux, D. D. Hodson, M. R. Grimaila, R. D. Engle, C. V. Mclaughlin, and G. B. Baumgartner. "Using Modeling and Simulation to Study Photon Number Splitting Attacks" In: IEEE Access 4, 2188. (2016) 10.1109/ACCESS.2016.2555759. |
| [Pirandola2020] | Pirandola, S.; Andersen, U. L.; Banchi, L.; Berta, M.; Bunandar, D.; Colbeck, R.; Englund, D.; Gehring, T.; Lupo, C.; Ottaviani, C.; Pereira, J. L.; Razavi, M.; Shaari, J. S.; Tomamichel, M.; Usenko, V. C.; Vallone, G.; Villoresi, P.; Wallden, P.. "Advances in quantum cryptography" (2020) 10.1364/AOP.361502. arXiv:1906.01645. https://opg.optica.org/aop/abstract.cfm?uri=aop-12-4-1012. |
| [Scarani2004] | V. Scarani, A. Acn, G. Ribordy, N. Gisin. "Quantum Cryptography Protocols Robust against Photon Number Splitting Attacks for Weak Laser Pulse Implementations" In: Phys. Rev. Lett. 92 (2004), p. 057901. (2004) 10.1103/PhysRevLett.92.057901. |
| [Zapatero2024] | Víctor Zapatero, Álvaro Navarrete, Marcos Curty. "Implementation Security in Quantum Key Distribution" (2024) 10.1002/qute.202300380. |
e6166f49-06a2-46a4-be5d-f06a99414837
- Quantum Communication
Technique properties
Operational properties
| Description | Countermeasure |
|---|---|
|
Decoy states are the operative defence against this attack. The transmitter randomly replaces signal pulses with pulses of different intensities, typically vacuum and one or more weaker states, and the catalogue anchor records that the most effective and widely known countermeasure against PNS attacks is the decoy-state method [BSI2023]. What the method pins down is the quantity the founding paper's security condition turns on, the multiphoton fraction against the fraction of pulses actually detected, since violating that condition is what hands the threat actor full information without inducing any errors or causing a change in the expected detection rate [Brassard2000]. On the same bench that ran the attack, with a signal mean photon number of 0.3 plus vacuum and one weaker decoy, the secure key rate computed by the GLLP formula turns to be negative under the threat actor's intervention [Liu2011]. The modern statement of why this settles the matter is that the detection statistics of the different intensities allow the single-photon detection statistics to be estimated tightly irrespective of what she does [Zapatero2024]. |
CM-0003 Decoy States Employment |
|
A true single-photon source removes the resource the attack consumes outright: with no multi-photon pulses there is no redundant copy to split, at any channel loss and against any protocol. That is the limiting case the whole technique is defined against, and in 2000 the founding paper stated that single-photon states had not yet been practically implemented for QKD although proposals existed and experiments have been performed to generate them for other purposes [Brassard2000]. Attenuated lasers are used precisely because this control is not available, which is why the weakness is the assumption that individual photons are being prepared [Zapatero2024]. |
CM-0002 True Single Photon Source |
|
Privacy amplification is what removes the threat actor's residual knowledge from the final key, and against this attack its usefulness turns entirely on whether that knowledge has been correctly bounded first. If her information is at most a given number of deterministic bits, a randomly chosen hash function leaves her expected information below a bound that shrinks exponentially in the number of bits discarded, and the legitimate parties estimate her information from the detected error frequency and the optical pulse intensity [Bennet1992]. This attack is the one that leaves the detected error frequency untouched, since each multi-photon signal will yield its complete information to an eavesdropper without leading to errors in the sifted key [Brassard2000] and the measured rate under attack was being the same as that of without attack [Liu2011], so an estimate built on error rate alone under-counts her and the compression is applied at the wrong ratio. The control is therefore a dependent step here rather than a standalone defence. |
CM-0001 Privacy amplification |
|
Employing another quantum communication protocol which is inherently PNS attack resistant. For example, using a modified BB84 protocol with non-orthogonal states [Scarani2004], a protocol family presented as robust against photon number splitting attacks for weak laser pulse implementations and which differs only in the classical sifting procedure [BSI2023]; or using a frequency coding scheme, which the anchor records can only partially inhibit the attack [Gabdulkhakov2018,BSI2023]. What each route removes is the property the founding paper's analysis turns on, that a multi-photon signal will yield its complete information to an eavesdropper without leading to errors in the sifted key [Brassard2000]. [Ashkenazy2024] names the three families invented against this attack, SARG04, decoy-state and coherent-one-way, and shows that its own implementation cannot beat coherent-one-way because the threat actor must reset the atom before each qubit, which completely destroys the coherence of the x-basis states and gives her away. The DPR-type route is now contested. A 2025 result reports a generalised photon-number-splitting attack combined with sequential unambiguous state discrimination which is efficient against the coherent-one-way protocol and introduces no error, where that protocol was previously considered secure against photon-number splitting: Avanesov and Kronberg, Physical Review A 111, 042601 (2025), doi 10.1103/PhysRevA.111.042601. A defender choosing a DPR-type protocol specifically to avoid this attack should treat that rationale as open rather than settled. |
CM-0006 Employing Another Quantum Protocol |
|
Incorporating this attack in the security proof and in the calculation of the secure key rate, leading to a proper amount of privacy amplification [Gottesman2004,BSI2023]. The leakage being accounted for is the one the founding paper derives, where each multi-photon signal will yield its complete information to an eavesdropper without leading to errors in the sifted key [Brassard2000]. What makes the control checkable rather than aspirational is that the accounting shows up as a number: evaluating the measured attack against the GLLP key-rate formula, the secure key rate turns to be negative under the threat actor's intervention, so the proof-level treatment surfaces exactly what the observed error rate does not [Liu2011]. The control is only as good as the leakage bound fed into it, which for this attack has to come from deliberately varied intensities rather than from the detected error frequency. |
CM-0014 Full Device Imperfections Incorporation in Scientific Security Proof |
| Description | Category | Impact Type |
|---|---|---|
|
Physical Layer (Quantum & Optical) is the primary target. The threat actor (Eve) exploits the multi-photon pulses emitted by the weak-coherent-pulse source and uses optical means (quantum non-demolition measurement, beamsplitter, quantum memory) to split and store photons while forwarding at least one to the receiver (Bob) [Brassard2000,Pirandola2020]. |
L1 — Physical Layer (Quantum & Optical) | Primary Target |
|
Post-Processing & Cryptographic layer is affected as a knock-on effect. The information Eve gains on the multi-photon pulses must be removed during privacy amplification, and the secure-key-rate computation (e.g. decoy-state estimation of the single-photon yield) must account for the PNS leakage [Pirandola2020,Gottesman2004]. |
L5 — Post-Processing & Cryptographic | Secondary Effect |
| Relation type | Description | Technique |
|---|---|---|
| Precedes |
The published framework's PNS chain follows this step with post-processing of the intercepted data, which turns what the threat actor learned into the final key [Krelina2025]. |
A-0040 Adversary Post-Processing |
| Relation type | Description | Technique |
|---|---|---|
| Preceded by |
Channel access is the step the photon-number-splitting attack starts from: the framework's worked PNS kill chain places a 'Tap fibre optic cable' step immediately before the PNS step, because the threat actor has to be on the line before she can split photons out of the pulses passing her position [Krelina2025]. |
A-0029 Compromise Quantum Channel |
| Preceded by |
A-0064 opens the door and A-0044 walks through it. Raising the mean photon number leaving Alice is what makes the multi-photon components large enough to split: the primary's strong model is explicitly the combined photon-number-splitting and cloning attack already assumed by Clavis2, but with the mean photon number manipulated and increased by a factor x so that the multi-photon components can no longer be ignored [Sajeed2015]. |
A-0064 Pulse-Intensity Increase in Two-Way QKD |
| Refined by |
This technique refines the Photon-Number-Splitting Attack by supplying it with an input the plain attack does not have: a per-pulse label saying whether that pulse was signal or decoy. Plain photon-number splitting is answered by the decoy-state method, because comparing the yields of two intensities bounds how much a splitter can have taken. Once the intensities are distinguishable by arrival time that bound stops holding, since the threat actor tunes separate observation windows to match the announced gain per intensity and so "the attack will remain unnoticed" [Huang2018]. |
A-0087 Timing Mismatch Attack on Signal and Decoy States |
| Has variant |
This technique is the wavelength-axis variant of the canonical photon-number-splitting attack A-0044. It shares the splitting stage exactly, picking photons out of the multiphoton pulses, attenuating the single-photon and multiphoton populations at different transmission rates and forwarding them over a lossless channel, and adds the one thing plain PNS lacks against a decoy-state system: a way to tell decoy pulses from signal pulses, supplied by the intensity modulator's rising-edge frequency shift. The dependency runs one way. Remove the labelling stage and what is left is conventional PNS, which is exactly what [Jiang2012] shows happens at a distinguishing error rate of 0.5, where the threat actor can get no information because the decoy-state method is used. |
A-0067 Wavelength-Selected Photon-Number-Splitting Attack |
| Has variant |
This technique is photon-number splitting with a discrimination stage in front: the unambiguous state discrimination labels each pulse signal or decoy so that the splitting stage is no longer bounded by the decoy-state estimator [Tang2013a]. |
A-0055 Non-Random Phase Attack |
|