Hacking quantum key distribution: implementation attacks

Short answer: Quantum key distribution (QKD) is secure within its theoretical bounds, but research papers show real devices can leak. A 2010 arXiv paper reported that the detectors in two commercial QKD systems could be remote-controlled with tailored bright light, letting an eavesdropper acquire the key. Later papers describe self-testing countermeasures and a testbench, and one reports that a pulsed blinding attack still evaded a simple countermeasure.

Theoretical Security vs. Physical Implementation

An AIP Publishing paper on countering detector manipulation attacks describes QKD as a communication method that uses quantum states of light as a trusted courier, so that any eavesdropping attempt is revealed by the underlying quantum physics of the measurement. It says that in practical systems, imperfect physical devices open security loopholes that challenge this core promise. The same paper says the basic protocols are secure within their theoretical bounds.

A 2010 arXiv paper says QKD implementations always rely on detectors to measure the relevant quantum property of single photons. A 2023 paper in EPJ Quantum Technology says the single-photon detector currently seems to be the most unsafe QKD system element, and that several attacks focus on it. The AIP Publishing paper says vulnerability of single-photon detectors to blinding attacks has been one of the biggest concerns, alongside various side channels.

Detector Manipulation via Bright Illumination

The 2010 paper reports experiments showing that the detectors in two commercial QKD systems could be fully remote-controlled using specially tailored bright illumination. The authors say this makes it possible to tracelessly acquire the full secret key, and they propose an eavesdropping apparatus built of off-the-shelf components. These are the authors' laboratory results from 2010 about the systems they tested. This article does not say how current systems fare.

The authors say the loophole is likely to be present in most QKD systems using avalanche photodiodes to detect single photons. They describe their findings as important for strengthening the security of practical QKD by identifying and patching technological deficiencies.

Countermeasures and Self-Testing

The AIP Publishing paper presents a countermeasure based on self-testing of detectors, to confirm their intended operation without relying on specific aspects of their inner working and to reveal manipulation attempts. The authors experimentally demonstrate it with a typical InGaAs avalanche photodetector and say it can be easily implemented with any single-photon detector. The paper says the problem has been addressed both by technical means and by advanced protocols.

The 2023 EPJ Quantum Technology paper reports an automated testbench that checks a detector's vulnerabilities against attacks using tailored bright illumination. It tested a free-running detector with a rudimentary countermeasure that measures an average photocurrent. The testbench automatically found the detector controllable in a continuous-blinding regime, and the countermeasure registered a photocurrent significantly exceeding that in a quantum regime, revealing the attack.

Automated Verification and Residual Risks

The authors then performed a pulsed blinding attack manually, which controls the detector intermittently. The countermeasure missed it in a wide range of blinding pulse durations and powers, still allowing the key to be eavesdropped. They make recommendations for improving both the testbench and the countermeasure. This article summarises abstracts and introductions of three papers and does not assess any product.

FAQ

Can bright light control QKD detectors?

A 2010 arXiv paper reports that detectors in two commercial QKD systems could be fully remote-controlled with specially tailored bright illumination, which would allow the key to be acquired without a trace. These were laboratory results on the systems tested.

Which detectors does the 2010 paper think are affected?

The authors say the loophole is likely present in most QKD systems that use avalanche photodiodes to detect single photons.

What is detector self-testing?

The AIP Publishing paper presents a countermeasure that checks a detector's intended operation without relying on its inner workings, demonstrated with an InGaAs avalanche photodetector.

Can automated testing catch every attack?

Not in the 2023 testbench study: a manually performed pulsed blinding attack was missed by the tested countermeasure over a wide range of pulse durations and powers.

Is the QKD protocol itself broken?

No. The AIP Publishing paper says the basic protocols are secure within their theoretical bounds. The concern is imperfect physical devices.