Quantum Entropy

QRNG · min-entropy · randomness extraction

Runs a modeled beam-splitter source into real entropy measurements, real von Neumann debiasing, and a real Toeplitz extractor, so you can watch “unpredictable by physics” get engineered into “uniform enough to be a key.”

What is this?

A quantum random number generator gets its unpredictability from physics: in the textbook description, a photon at a beam splitter has not decided which detector it will hit until it is measured. But real hardware is imperfect — one detector is slightly more sensitive, clicks echo into the next time slot, and the raw bits come out lopsided and streaky. This lab is about the gap between random and usable: how to honestly measure how much surprise a stream really contains, and how to concentrate it into fewer, perfectly balanced bits that could actually serve as a key. Not production crypto — a teaching demo.

Where this sits

The real pipeline — and what this lab is not

A deployed randomness stack (NIST SP 800-90C) chains an entropy source through continuous health tests and a conditioner into a deterministic bit generator. This lab is the front half; the siblings own the rest.

Entropy sourcethis lab · modeled QRNG Health teststhis lab · SP 800-90B RCT + APT Conditioning / extractionthis lab · Toeplitz DRBG seedentropy-collapse DRBGdrbg-arena · corrupted-oracle

What this isn’t

Not a QKD lab. Quantum measurement used for key agreement lives in bb84 and e91; here measurement only generates local randomness.
Not a DRBG lab. Everything after the seed — including the SP 800-22 statistical suite — belongs to drbg-arena. Note in passing: a von-Neumann-debiased correlated stream passes plenty of statistical tests while being predictable — which is exactly why this lab measures predictability instead.
Not an entropy-failure lab. entropy-collapse shows what happens when a seed is shared; this lab is about a seed that is weak.
No real quantum hardware, and no device-independent build-out. The source here is a labeled model. Device-independent randomness — certified by a Bell-test violation (see e91’s CHSH exhibit) — is the only construction that does not require trusting the manufacturer. It exists, it is slow, and it is rare; we name it rather than build it.