132
Part Three · C · Complexity and cryptography

Quantum key distribution: BB84, E91, security proofs, device independence

Every secure website you use agrees a secret key with mathematics that Shor’s algorithm would break. Quantum key distribution (QKD) offers something different: a key whose secrecy follows from the laws of physics, because measuring an unknown quantum state disturbs it, and the disturbance can be detected. The theory is beautiful and the proofs are rigorous. Real systems have nonetheless been hacked, not through the physics but through the hardware that implements it.

On the plate: you are Eve. Try to read Alice’s key without Bob noticing.

BB84

Charles Bennett and Gilles Brassard proposed the first protocol in 1984. Alice sends single photons, each encoding a random bit in a randomly chosen basis: rectilinear (0° or 90°) or diagonal (45° or 135°). Bob measures each in a random basis. Afterwards they announce bases, not bits, over a public channel and keep only the rounds where the bases matched: the sifted key, about half the photons. They sacrifice a sample to estimate the quantum bit error rate (QBER). By no-cloning (Chapter 102), Eve cannot copy the photons; any measurement in the wrong basis randomizes them. Established

Play Eve

Intercept a fraction 0.50 of the photons, measure each in a random basis, and resend what you saw. Then . Half the time you guess the wrong basis; then Bob’s result is random, wrong half of those times. Intercepting everything gives a QBER of 12×12=25%\tfrac12\times\tfrac12 = 25\%. Below a QBER of 11%, error correction and privacy amplification (hashing the key down to a shorter one that Eve knows almost nothing about) still leave a secret key at rate 1−2h(Q)1 - 2h(Q) per sifted bit, where hh is the binary entropy. Peter Shor and John Preskill proved this in 2000 against any attack Eve can make, by relating BB84 to an error-corrected entanglement-distillation protocol. Real keys also need finite-size corrections, reviewed by Christopher Portmann and Renato Renner in 2022. Established

Hacking the hardware, and device independence

In 2010 Lars Lydersen and colleagues blinded the single-photon detectors of commercial QKD systems with bright light, turning them into classical detectors that Eve could control remotely. She obtained the full key with no rise in QBER. The proofs assumed devices behaved as modelled; they did not.

Artur Ekert’s 1991 protocol uses entangled pairs and a Bell test (Chapter 51). Its descendant, device-independent QKD, certifies security from the observed CHSH value SS alone, treating the devices as black boxes. With a QBER of 0.020 the plate plots the key rate 1−h(Q)−h(1+(S/2)2−12)1 - h(Q) - h\big(\tfrac{1+\sqrt{(S/2)^2-1}}{2}\big). In 2022 David Nadlinger and colleagues at Oxford ran it with trapped ions, distilling 95,628 key bits from 1.5 million Bell pairs. In 2026 Bo-Wei Lu, Jian-Wei Pan and colleagues at USTC reported in Science device-independent QKD between two single-atom nodes linked by 100 km of fibre. Whether QKD is worth deploying, given its cost and the need for trusted relays, rather than post-quantum cryptography (Chapter 133), is argued among security agencies. Debated

What you now have

BB84 and sifting, QBER, the intercept–resend attack and its 25% signature, the Shor–Preskill bound and its 11% threshold, privacy amplification, detector blinding, E91 and device-independent QKD with its key-rate formula.

· · ·

Three questions to carry with you.

Why does intercept–resend cause 25% errors?tap
Eve picks the wrong basis half the time, and then Bob’s bit is wrong half the time.
What does privacy amplification do?tap
Hashes the corrected key into a shorter one about which Eve has negligible information.
What does device independence remove?tap
Trust in the devices’ inner workings; security follows from the observed Bell violation.

And one to work out. Eve intercepts and resends every photon in a random basis. What QBER results?

QBER =

Sources: C. H. Bennett and G. Brassard, Proc. IEEE Int. Conf. Computers, Systems and Signal Processing, 175 (1984). A. K. Ekert, Phys. Rev. Lett. 67, 661 (1991). P. W. Shor and J. Preskill, Phys. Rev. Lett. 85, 441 (2000). L. Lydersen et al., Nat. Photon. 4, 686 (2010). S. Pironio et al., New J. Phys. 11, 045021 (2009). D. P. Nadlinger et al., Nature 607, 682 (2022). C. Portmann and R. Renner, Rev. Mod. Phys. 94, 025008 (2022). T. Vidick and S. Wehner, Introduction to Quantum Cryptography (CUP, 2023). B.-W. Lu et al., Science 391, 592 (2026). Checked by code: 1 − 2h(Q) reaches zero at Q = 0.1101; DI rate at S = 2.7, Q = 0.02 is 0.587.

Next · Chapter 133
Post-quantum cryptography: lattice, code, and hash-based schemes, plus standardization

The real defense against Shor's algorithm is new mathematics, not quantum hardware.

Turn the page