Power Consumption Analysis Attack on State Preparation in QKD Systems
The threat actor (Eve) infers the quantum states prepared by the transmitter (Alice) by analysing the power consumption of the electrical circuit that drives state preparation, using machine learning [Zheng2021] or template matching [Dacosta2025] to correlate power with the transmitted states. The circuit can be an integrated electrical control circuit on a transmitter chip, which is where the attack was first analysed [Zheng2021], or a discrete electronic driver such as a field-programmable gate array board, which is where it was first measured [Dacosta2025].
What Eve walks away with is part of the key rather than all of it, a measured 73.35 percent accuracy in predicting the transmitted qubits in the demonstrated case [Dacosta2025]. The attack is invisible to the link's own checks, because it reads an electrical quantity and never touches the light, so the transmitted states, the receiver's (Bob's) measurements and the error rate the two parties estimate are all unchanged [Zheng2021]. Alice and Bob see a healthy link and claim a key that Eve already knows part of.
| [BSI2023] | BSI. "Implementation Attacks against QKD Systems". (2023) https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Publications/Studies/QKD-Systems/QKD-Systems.pdf. |
| [Dacosta2025] | Beatriz Lopes da Costa, Matias R. Bolanos, Ricardo Chaves, Claudio Narduzzi, Marco Avesani, Davide Giacomo Marangon, Andrea Stanco, Giuseppe Vallone, Paolo Villoresi, Yasser Omar. "Power-consumption back door in quantum key distribution" (2025) 10.1103/f92x-c3zj. arXiv:2503.11767. https://link.aps.org/doi/10.1103/f92x-c3zj. |
| [ETSI008] | ETSI. "Quantum Key Distribution (QKD); QKD Module Security Specification". (2010) https://www.etsi.org/deliver/etsi_gs/QKD/001_099/008/01.01.01_60/gs_QKD008v010101p.pdf. |
| [Li2026] | Zijian Li, Chenyu Xu, Xin Hua, Yongqiang Du, Xin Liu, Tao Lin, Xi Xiao, Kejin Wei. "Security risks of VOA-induced luminescence in chip-based quantum key distribution" (2026) 10.1038/s41534-026-01365-1. |
| [Zheng2021] | Y. Zheng, H. Shi, W. Pan, Q. Wang, and J. Mao. "Quantum Hacking on an Integrated Continuous-Variable Quantum Key Distribution System via Power Analysis" In: Entropy 23, 176.. (2021) 10.3390/e23020176. |
8c9153d7-d1e6-4df5-9605-0b02c4dc2b76
- Quantum Communication
Technique properties
Operational properties
| Description | Countermeasure |
|---|---|
|
Power Analysis Masquerading and Obscuring is the control the primary source recommends against this attack by name, and three of its six measures reach it. Randomising the circuit's power is the measure that attacks the mechanism, because "the electrical control circuit can be improved by randomizing the power to close this loophole", and the second is that "dynamic voltage and frequency scaling (DVFS) technology can be applied to reduce the dynamic power" [Zheng2021]. The third is that "the pipeline structure and parallel structure can be adopted to optimize the electrical control circuit to reduce the power" [Zheng2021]. The other three measures, randomisation of syndrome-computation order, insertion of dummy operations and secret permutation rules, are reconciliation-phase measures with nothing to act on at state preparation, so the control is scoped rather than wholly applicable here. ETSI GS QKD 008 clause 4.8 makes this more than a recommendation, since "The QKD module shall protect the CSPs against SPA attacks." and the same clause requires protection against differential power analysis [ETSI008]. One caveat: the support is mechanism-level, the author's claim that randomisation closes the loophole, and not a measured bound, since no source reports an attack strength after any of these measures [Zheng2021]. |
CM-0070 Power Analysis Masquerading and Obscuring |
|
Tamper Protection denies the access this attack cannot proceed without. The threat actor has to reach the driver's supply inside the module, and such attacks "have rarely been used in QKD, since they require an eavesdropper to have access to Alice's or Bob's setup" [Dacosta2025]. CM-0076's own steps implement exactly the perimeter that stands in her way, since they define the cryptographic boundary and protect everything inside it, use an opaque production-grade enclosure with locks or numbered seals, apply a removal-resistant coating or a tamper-detection envelope, and run tamper-response circuitry that zeroizes the critical security parameters on detection [ETSI008]. Unlike CM-0070, it does not touch the leak itself, it stops the threat actor getting to it. |
CM-0076 Tamper Protection |
|
A passive transmitter removes the quantity this attack reads. CM-0034's own text is that the design avoids the need for active modulation elements, replacing them with a linear optics network and post-selection, and the leak exists only because "for the encoding of different key information, the required modulation voltages are different" [Zheng2021]. A transmitter with no key-dependent drive voltage has nothing for the threat actor to correlate, which is architectural elimination rather than the signal suppression CM-0070 provides. |
CM-0034 Passive Quantum Communication Transmitters |
| Description | Category | Impact Type |
|---|---|---|
|
The Physical Layer is a prerequisite the threat actor must already hold, not a layer this attack affects. Nothing optical is disturbed, which is the row's own basis for qber_impact none, but reaching the driver's supply means already being inside the module's physical envelope, which ETSI GS QKD 008 calls "an explicitly defined continuous perimeter that establishes the physical bounds of a QKD module" [ETSI008] and which the version of record names as the practical barrier, since such attacks "have rarely been used in QKD, since they require an eavesdropper to have access to Alice's or Bob's setup" [Dacosta2025]. The catalogue already places that envelope at L1 through the mapping of CM-0076 Tamper Protection, the control linked on this row for exactly this reason. |
L1 — Physical Layer (Quantum & Optical) | Prerequisite for Attack |
|
Electronics and Analog Components is the primary target, because what the threat actor reads is the supply power of an electrical circuit and nothing optical takes part in producing it or in reading it [Zheng2021] [Dacosta2025]. The sibling row A-0095 maps its own power and emanation route to L2 on the same reading, that what leaks is from electronics rather than from light, and ETSI GS QKD 008 groups power analysis with the other non-invasive electrical side channels in clause 4.8 [ETSI008]. |
L2 — Electronics & Analog Components | Primary Target |
| Relation type | Description | Technique |
|---|---|---|
| Alternative to |
A-0095 Information Leakage Through Electromagnetic Radiation reads the same electronics by the other coupling path. A-0095 picks up radiated emission at a standoff, which is why its controls are shielding and TEMPEST, while this technique reads the conducted supply current at the rail, which shielding does not reduce [Zheng2021] [Dacosta2025]. ETSI GS QKD 008 clause 4.8 treats them as separate obligations, requiring protection against simple power analysis and differential power analysis and electromagnetic emanation in three distinct sentences, the last being "The QKD module shall protect the module's CSPs against EME." [ETSI008]. An operator who has done one has not done the other, and the two rows should be assessed together. |
A-0095 Information Leakage Through Electromagnetic Radiation |
|