Jorge Quitério Search
IT & Engineering

Quantitative limits of quantum key distribution

Quantum key distribution derives its security from physics rather than computational assumptions. Its limits are nevertheless quantitative: an 11% error threshold, exponential photon loss with distance, and a classical channel that still requires authentication. A BB84 simulation with an eavesdropper illustrates them.

Two quite different things are called quantum cryptography. Post-quantum cryptography is classical mathematics designed to resist quantum computers; I described its core idea in learning with errors and post-quantum key exchange. Quantum key distribution (QKD) uses quantum physics itself to agree a secret key, with security that does not depend on any computational assumption.

Its central claim is that the laws of physics, rather than the difficulty of a mathematical problem, protect the key. Its practical limits can be read from three formulas. This article presents those formulas and a simulated fibre link with an eavesdropper, Eve, showing how she is detected.

The relevant literature is as follows.

Bennett and Brassard (1984) proposed BB84, the first QKD protocol.

Wootters and Zurek (1982) proved the no-cloning theorem: an unknown quantum state cannot be copied. It is what makes eavesdropping detectable, and also what forbids amplifying quantum signals.

Shor and Preskill (2000) gave a simple security proof of BB84 and its secret-key rate, 1−2h(Q)1 - 2h(Q).

Lo, Curty and Qi (2012) proposed measurement-device-independent QKD, which removes attacks on detectors.

Liao and colleagues (2017) reported satellite-to-ground QKD from the Micius satellite over distances of up to about 1,200 km.

ANSSI, BSI, NLNCSA and the Swedish Armed Forces (January 2024) published a joint position paper concluding that QKD has significant functional limitations, suits niche use cases, and that priority should go to post-quantum cryptography.

BB84 in brief

Alice sends single photons, each encoding a random bit in one of two randomly chosen bases, + or ×. Bob measures each in a basis he also chooses at random. Afterwards they announce, over an ordinary public channel, which bases they used (not the bits) and keep only the positions where the bases matched: on average half, the sifted key.

An eavesdropper who intercepts a photon must measure it without knowing the basis. Half the time she picks the wrong one and disturbs the photon; in half of those cases Bob then reads the wrong bit. Intercepting every photon therefore introduces errors in about 12×12=25%\tfrac12 \times \tfrac12 = 25\% of the sifted bits. Eavesdropping therefore appears as errors.

The principle resembles a tamper-evident seal on a parcel. The seal does not prevent the parcel from being opened, but an opened parcel cannot be resealed without visible signs, so the recipient knows whether to trust the contents. In QKD the seal is a property of physics, and the visible sign is the error rate.

Estimating the error rate is a sampling problem

Alice and Bob estimate the quantum bit error rate QQ by publicly comparing a random subset of the sifted key, then discarding it. This is the deviation-rate problem of statistical sampling in control testing. With 10,000 test bits and an observed 3%, the standard error is

0.03×0.9710 000≈0.17%, \sqrt{\frac{0.03 \times 0.97}{10\,000}} \approx 0.17\%,

so the 95% interval is roughly ±0.33 percentage points. Finite-key security proofs account for this uncertainty, and for short keys the cost is significant.

The 11% threshold

Errors also come from imperfect detectors and channels, so the question is how much error is tolerable. After error correction and privacy amplification, which compresses the key to remove whatever Eve might know, Shor and Preskill's rate per sifted bit is

r=1−2 h(Q),h(Q)=−Qlog⁡2Q−(1−Q)log⁡2(1−Q). r = 1 - 2\,h(Q), \qquad h(Q) = -Q\log_2 Q - (1-Q)\log_2(1-Q).

One h(Q)h(Q) accounts for error correction, the other for privacy amplification.

QQ 1% 3% 5% 8% 10%
Secret fraction rr 0.84 0.61 0.43 0.20 0.06

It reaches zero at Q≈11.0%Q \approx 11.0\%. Above that, the protocol aborts.

Distance and channel loss

Fibre loses about 0.2 dB per kilometre at telecom wavelengths, so the fraction of photons surviving dd km is

T(d)=10−0.2 d/10: T(d) = 10^{-0.2\,d/10} :

1% at 100 km, 10−610^{-6} at 300 km, 10−1010^{-10} at 500 km. A direct fibre of 200 km, roughly the distance between Luxembourg and Brussels, would deliver about one photon in ten thousand; a direct link between Luxembourg and Lisbon would deliver practically none. Classical signals overcome loss with amplifiers. Quantum signals cannot be amplified, because of the no-cloning theorem. The two ways around this limit are trusted nodes, which relay the key hop by hop but hold it in the clear, and satellites, whose path runs mostly through near-vacuum. Europe's EuroQCI initiative plans both terrestrial and space segments.

Requirements outside the physical layer

QKD guarantees the key is secret. It does not say who is at the other end. The public channel must be authenticated, otherwise an attacker can run the protocol separately with each side. Authentication needs pre-shared keys or classical signatures, and for long-term security those signatures must be post-quantum. Real sources sometimes emit more than one photon and real detectors can be manipulated; decoy states and measurement-device-independent protocols address some of these attacks, at a cost in complexity.

The simulation below illustrates three results: an eavesdropper who intercepts a fraction ff of the photons adds about f/4f/4 to the error rate; above about 11% no secret key survives; and the photon count decreases exponentially with distance.

The simulation implements BB84 with random bases, an intercept-and-resend Eve, equipment errors and fibre loss.

Suggested settings:

  1. No eavesdropper. With no Eve, 2% equipment error and 50 km, about 10% of photons arrive, half of those survive sifting, QQ is about 2% (give or take half a point of sampling noise), and roughly two thirds of the sifted bits become secret key.
  2. Full interception. Set Eve's share to 100%. QQ jumps to about 25%, far above 11%, and the protocol aborts. The eavesdropper gains no usable key material and is clearly detected.
  3. Partial interception. Set the share to 20%, 30% and 40%. Each 10% of interception adds about 2.5 points of error. Somewhere near 35-40% the error rate crosses 11% and the key disappears; below that, privacy amplification removes her partial information, and the key becomes shorter.
  4. Distance. With no Eve, increase the fibre length to 200 km and then 300 km. The arriving photons fall from thousands to a handful to none. No protocol design overcomes this exponential loss.

Derivation of f/4. A photon Eve intercepts is measured in the wrong basis with probability 1/2. If so, she resends it in her basis, and Bob, measuring in Alice's basis (only those positions are kept), gets a random result: wrong with probability 1/2. So each intercepted sifted bit is wrong with probability 1/4, and intercepting a share ff adds f/4f/4 to QQ. Combined with an equipment error e0e_0, Q≈e0+f/4Q \approx e_0 + f/4 when both are small. The key rate 1−2h(Q)1 - 2h(Q) is zero when h(Q)=1/2h(Q) = 1/2, which happens at Q≈0.110Q \approx 0.110. For the loss, 0.2 dB/km means a factor of 10−0.0210^{-0.02} per kilometre, so T(d)=10−0.02dT(d) = 10^{-0.02 d}.

Practical implications:

  • Post-quantum cryptography is the migration path: it runs on existing networks and hardware.
  • QKD is a specialised complement for links where dedicated infrastructure exists and the threat model justifies it, and its authentication still relies on classical, preferably post-quantum, cryptography.
  • The useful numbers are simple: 11% error, 0.2 dB per kilometre, and the fact that physics does not remove the need to authenticate the other party.

References

  1. W. K. Wootters and W. H. Zurek, "A single quantum cannot be cloned", Nature, 299, 802-803, 1982.
  2. C. H. Bennett and G. Brassard, "Quantum cryptography: public key distribution and coin tossing", Proceedings of the IEEE International Conference on Computers, Systems and Signal Processing, Bangalore, 175-179, 1984.
  3. P. W. Shor and J. Preskill, "Simple proof of security of the BB84 quantum key distribution protocol", Physical Review Letters, 85(2), 441-444, 2000.
  4. H.-K. Lo, M. Curty and B. Qi, "Measurement-device-independent quantum key distribution", Physical Review Letters, 108, 130503, 2012.
  5. S.-K. Liao et al., "Satellite-to-ground quantum key distribution", Nature, 549, 43-47, 2017.
  6. ANSSI, BSI, NLNCSA and Swedish Armed Forces, Position Paper on Quantum Key Distribution, January 2024.