Saturation Attack
The threat actor (Eve) resends coherent states with a displaced mean to the CV-QKD receiver (Bob) to saturate the detector, masking the increase of excess noise from an intercept-and-resend attack [Qin2016].
The attack is possible because of what CV-QKD does not look at. Security is judged from the spread of the receiver's measurements and never from their average, so the threat actor is free to shift that average [Qin2013,Qin2016]. Shifting it far enough pushes part of the measurement outside the range where the detector still responds proportionally, and the clipped readings make the link look quieter than it really is.
The outcome is not a weakened link but a broken one. The transmitter and receiver accept a key the threat actor can recover in full, at a rate they believe is secure, and nothing in their own checks tells them otherwise [Qin2013,Qin2016].
The attack also reaches past optical fibre. A simulation study extends it to an open-air microwave CV-QKD link, because what it exploits is the receiver's own response rather than anything about the fibre [Peng2026].
| [ETSI016PP] | ETSI. "Quantum Key Distribution (QKD); Common Criteria Protection Profile - Pair of Prepare and Measure Quantum Key Distribution Modules". In: ETSI GS QKD 016. (2024) https://www.etsi.org/deliver/etsi_gs/QKD/001_099/016/02.01.01_60/gs_QKD016v020101p.pdf. |
| [ISOIEC23837P2] | ISO/IEC JTC 1/SC 27. "ISO/IEC DIS 23837-2:2022 Information security - Security requirements, test and evaluation methods for quantum key distribution - Part 2: Evaluation and testing methods". (2022) |
| [Kumar2021] | Rupesh Kumar, Francesco Mazzoncini, Hao Qin, and Romain Alleaum. "Experimental Vulnerability Analysis of QKD Based on Attack Ratings" In: Scientific Reports 11, 9564. (2021) 10.1038/s41598-021-87574-4. |
| [KunzJacques2015] | S. Kunz-Jacques and P. Jouguet. "Robust shot-noise measurement for continuous-variable quantum key distribution" In: Phys. Rev. A 91, 022307. (2015) 10.1103/PhysRevA.91.022307. |
| [Mao2020a] | Y. Mao, W. Huang, H. Zhong, Y. Wang, H. Qin, Y. Guo, and D. Huang. "Detecting quantum attacks: a machine learning based defense strategy for practical continuous-variable quantum key distribution" In: New J. Phys. 22, 083073. (2020) 10.1088/1367-2630/aba8d4. |
| [Peng2026] | Wenqi Peng, Xuan Li, Yongxin Feng, Xiu-Bo Chen, Ying Guo. "Saturation attack of the microwave-propagated continuous variable quantum key distribution in open air" (2026) 10.1088/1367-2630/ae856d. https://doi.org/10.1088/1367-2630/ae856d. |
| [Qin2013] | Hao Qin, Rupesh Kumar, and Romain All ́eaume. "Saturation attack on continuousvariable quantum key distribution system" In: Proceedings Volume 8899, Emerging Technologies in Security and Defence; and Quantum Security II; and Unmanned Sensor Systems X; 88990N. (2013) 10.1117/12.2028543. |
| [Qin2016] | H. Qin, R. Kumar, and R. Allaume. "Quantum hacking: Saturation attack on practical continuous-variable quantum key distribution" In: Phys. Rev. A 94, 012325. (2016) 10.1103/PhysRevA.94.012325. |
| [Xu2022] | S. Xu, Y. Li, Y. Mao, and Y. Guo. "Counteracting a Saturation Attack in Continuous-Variable Quantum Key Distribution Using an Adjustable Optical Filter Embedded in Homodyne Detector" In: Entropy 24, 383. (2022) 10.3390/e24030383. |
b8be7c79-87fc-453c-ad3a-8d54347cd4f3
- Quantum Communication
Technique properties
Operational properties
| Description | Countermeasure |
|---|---|
|
A variable optical attenuator in the receiver's signal path, driven by feedback from the detection response, keeps the received quadratures inside the linear range so the clipping never happens. [Xu2022]'s adjustable element performs an actual gain adjustment based on the feedback of the response of detection, attenuating each block according to a coefficient recomputed from the previous one, and with it applied the estimated excess noises fall back into the finite linearity domain. It performs no spectral, spatial or polarisation selection, which is why it is a VOA rather than a filter. With the masking removed, the corrected estimate gives no secret key and the parties detect the attack [Xu2022]. |
CM-0099 Variable Optical Attenuator |
|
Acquiring quadrature measurements in real time at randomly chosen optical attenuations tests the linearity the security proof assumes, and clipping destroys that linearity. [KunzJacques2015] simulated a saturation attack against the method: the squared correlation coefficient of the noise-to-signal fit fell to 0.386 at receiver variance 20 and 0.665 at variance 1, against an acceptance threshold of 0.99. [Qin2016] independently calls it precisely the countermeasure that constitutes an effective parade against the saturation attack. One configuration detail decides whether it works at all. More than two attenuation values are required, because with the usual two-point scheme the signal is shut off to measure shot noise and the saturated signal is blocked along with it, so the shot-noise figure is unaltered and a simple excess-noise difference may be defeated [KunzJacques2015]. [Qin2016] states the same boundary from the other side: the attack cannot be performed if the total noise variance is estimated with more than two attenuation values and the linearity of the total noise against attenuation is checked. |
CM-0087 Real-time Quadrature Measurements with Random Optical Attenuations |
|
A neural network trained on pulse-level features detects this attack without being told to look for it, and the saturation attack turns out to be among the easiest of the six classes it handles. [Mao2020a] finds that the feature vectors of the saturation attack are very different from the normal unattacked vectors, where those of the local-oscillator intensity attack are close to them and hard to separate by statistical analysis. Precision and recall for the saturation attack both reach the maximum of 1 at fifteen hidden-layer neurons, with false-positive and false-negative rates reaching 0 at the same point [Mao2020a]. The evidence level is simulation, and the scheme carries a standing cost: 10 percent of pulses are spent estimating the shot-noise variance and the amplitude modulator in Bob's signal path adds a typical 2.7 dB of insertion loss [Mao2020a]. A line of work has continued since, through quantum k-nearest-neighbour and hybrid quantum-classical classifiers between 2024 and 2026, including a countermeasure proposed alongside the free-space extension of this attack [Peng2026]. |
CM-0088 Pattern Recognition Using Machine Learning |
|
Discarding whole measurement blocks that contain any reading outside the detector's confidence interval stops the bias forming, because clipped data never reaches parameter estimation. This is the control the attack's own primary source recommends: [Qin2016] proposes discarding measurement data blocks if any of the measured data falls out of the confidence interval and states that this counter-measure will efficiently protect against saturation attacks and is easily implementable, with the interval typically larger than six standard deviations so that the probability per measurement event of a saturation is below 2e-9. [Kumar2021] describes it, citing [Qin2016], as the option that needs no additional hardware, relies on pre-characterisation of the detector linearity range and has a small marginal cost. The obvious implementation is the wrong one. A plain confidence-interval post-selection that discards individual points rather than whole blocks would give the threat actor the power to influence which data is post-selected just by controlling the displacement value, leaving the retained data non-Gaussian with no security guarantee [Qin2016]. The control must therefore be block-wise. A refined variant, Gaussian post-selection, allows some key to be generated even under moderate saturation, retaining 15.37 percent of points in simulation at an L2 distance to a perfect Gaussian below 1e-3 [Qin2016]. ISO/IEC 23837-2 clause 8.10 raises this from a published proposal to the pass condition of a drafted international test standard: the receiver module under test must reject the data or report an exceptional event once the homodyne detector operates in the saturation region, and must abandon data recorded in that region [ISOIEC23837P2]. |
CM-0089 Discarding Non-linear Data Blocks |
|
A watchdog photodiode at the receiver's optical input raises an alarm on light the link should not be receiving. [Qin2016] proposes it against this attack: the displacement must be strong, so the signals reaching the homodyne detection carry a high energy, which an auxiliary detector can upper-bound. Scope limits matter here. [Qin2016] warns that it is in general not possible to predict in which mode an attacker will try to send energy in order to saturate the detector, so it can be very difficult to design in practice a watchdog capable of detecting any such attempt. ISO/IEC 23837-2 clause 8.5.1(e) states the same limit normatively, that there is some minimum power below which a dedicated monitor is not sensitive and powers lower than that shall still be tested [ISOIEC23837P2]. |
CM-0011 Watchdog Detector |
| Description | Category | Impact Type |
|---|---|---|
|
Physical quantum and optical is the layer the attack arrives through. The saturating light is delivered into the receiver's signal port as a displacement of the re-sent states [Kumar2021]; a simulation study extends the attack to an open-air microwave link [Peng2026]. Secondary rather than primary because no optical component is itself defective, the optics only carry the power that drives the electronics out of range. |
L1 — Physical Layer (Quantum & Optical) | Secondary Effect |
|
Post-processing and cryptographic is where the attack's effect is realised. The clipped readings are consumed by parameter estimation, which uses only second-order moments and never the mean, so the transmission and excess-noise estimates come out biased and the secret-key-rate formula is computed from wrong inputs [Qin2016]. Secondary because the arithmetic at this layer is correct and only its inputs are corrupted. It matters for defence because the two most effective controls act here rather than optically, by discarding data recorded in the saturation region [Qin2016] or by testing the signal-to-noise linearity across random attenuations [KunzJacques2015]. |
L5 — Post-Processing & Cryptographic | Secondary Effect |
|
Electronics and analog components is the layer the attack primarily hits, and the sources are unusually specific about it. [Qin2016] aims the attack at the homodyne detection on Bob's side and more specifically at the electronics of the homodyne detection, and [KunzJacques2015] attributes the bounded response to limitations of the analog electronics of the detection or to the range of the analog-to-digital conversion. The limits that gate the attack are voltages at an amplifier output and a digitiser's configured range [Kumar2021,Qin2016], not optical powers. |
L2 — Electronics & Analog Components | Primary Target |
| Relation type | Description | Technique |
|---|---|---|
| Alternative to |
This row holds the coherent-displacement route, in which the threat actor resends displaced coherent states that saturate the receiver's homodyne detector [Qin2016]. A-0102 CV Detector-Control Attack holds the incoherent-injection route, in which a separate laser shines into the receiver's signal port [Kumar2021]. [Kumar2021] treats the two as strategies of one saturation attack, and [Mao2020a] lists them as two separate attacks. [Kumar2021] attributes its own incoherent route to the paper behind A-0102 and shows the choice is economic rather than mechanical, Attack Potential 14 against 26, because an incoherent laser needs no phase-drift compensation [Kumar2021]. |
A-0102 CV Detector-Control Attack |
|