I had once heard that trng (true random generator), which CPU vendors and security dongles manufacturers like to advertize and put forward as a true source of randomness, was a bit overrated a term because really true randomness can only be produced out of a device based on quantum mecanics properties.
And it looks like some scientists have just produced a certifiably perfect source of randomness using entangled superconducting qubits in an experiment:
Interesting stuff! For those of you who understand quantum mecanics (who does, really?), here are some more details:
The experiment by ETH Zurich’s Department of Physics amplified imperfect randomness into mathematically certifiable perfect randomness using a high-precision Bell test, a quantum physics experiment used to verify entanglement between particles
The experiment used two superconducting quantum chips connected through a 30-meter cryogenically cooled link that allowed microwave photons to travel between them. The photons created quantum entanglement between the qubits, enabling measurements performed on one qubit to instantly influence the state of the other
I think current devices are pretty solid because underlying their operation they probably do rely on some quantum process. You may lose a few bits of entropy due to biases but most of the time your security does not depend on those last few bits.
However I prefer not to rely on the “rand” mechanism built in to a CPU (or built in to an OpenPGP card) since you have exactly zero visibility as to how the sequence of numbers arises. For the same reason though
Public randomness services
doesn’t make a lot of sense to me. There are already public randomness services but for cryptographic purposes, or anything else of high importance, that is not really safe. You have no visibility as to whether they are really generating the randomness how they say they are and you don’t know whether they record the randomness that is handed out even if it is good randomness.
And, as always, crypto security is only as good as its weakest link … so the randomness part of it by itself needs to be analysed for threats / attacks / weaknesses from start to finish (never mind about the rest of the process).
