Skip to content

Quantum random number generators: the complete guide

A random number generator produces values that cannot be predicted by an observer who knows how the generator is built. Cryptography depends on that unpredictability more than on any algorithm choice: every key, nonce and initialisation vector is only as strong as the randomness it came from.

35 articles in 5 sections

Almost every failure people call a "crypto bug" is really a randomness bug. The algorithm did what it was specified to do, and the number fed into it was more predictable than anyone realised. That is why this section starts with the generator rather than the cipher.

The families, and why the distinction matters

Four terms get used interchangeably and should not be. A PRNG expands a seed deterministically. A CSPRNG does the same but is built so that observing its output does not let you predict the rest. A TRNG derives unpredictability from a physical process. A QRNG is a TRNG whose physical process is quantum, which means its unpredictability rests on a physical model rather than on an assumption that the process is too messy to model.

The practical consequence is not that one family is good and another bad. It is that they fail differently, and the evidence you can produce for each is different. The comparison is worth doing carefully, because a correctly seeded CSPRNG is entirely adequate for a great many systems.

Where the hard question actually is

Choosing a generator family is the easy part. The hard question is what evidence you have that the source is still behaving the way it did when it was characterised — not at the factory gate, but now, in the system you are running. That question is what the entropy pillar exists to answer, and it is where this section hands off.

No generator proves its own output is random, and nobody can make that claim. What can be established is a physical model of the source, an estimate of min-entropy conditional on that model, and runtime evidence that the source has not drifted away from it.

Where to start

If the terms are new, begin with the fundamentals below. If you want to know how a quantum source actually works, the technology section explains phase diffusion in enough detail to be checked. The maturity model gives you a way to grade any source, including ours — and the deployment and decision sections are for the point where you stop reading and start specifying.

Fundamentals

What a random number generator is, how the four families differ, and which one a given system actually needs.

How the technology works

Inside a quantum entropy source — the physics, the architectures compared, and what happens between the laser and the bits.

The four levels of RNG

A way to grade any source, and what separates one that was tested once from one you can check while it is running.

Deployment and integration

Throughput, form factors, driver integration and the HSM case — the questions that arrive once you start specifying.