hap.random: a new random number library for D

[Nick Sabalausky via Digitalmars-d-announce]

On 6/11/2014 12:35 PM, Kagamin wrote:
>
> In some scenarios impredictability is not enough. For example, when you
> generate a session id, an attacker doesn't have to predict it ahead of
> time, he can guess it at any time later. And if they listen to radio
> waves - that's an "open protocol", an attacker can setup antenna near
> their antenna and get the same readings.

An interesting point.

> Cryptographic PRNG and quantum
> TRNG are better isolated, so it's harder to read them.

FWIW, a cryptographic PRNG isn't necessarily well-isolated. Being a 
PRNG, the isolation of a cryptographic PRNG is primarily limited to two 
main things:

- The isolation of its entropy source(s) (which are not normally part of 
a crypto-PRNG's specification - it's just left as "choose a good one"), and

- The patterns of how data is drawn from the PRNG.

If the entropy source is poorly isolated (via poor choice of entropy 
source, or a failure within the entropy source), and the requests being 
made to the PRNG are relatively predictable or even guessable (quite 
likely given the nature of software), then a cryptographic PRNG won't be 
any better isolated than, say, the digits of PI.

TL;DR: The isolation of a cryptographic PRNG is that of its external 
entropy source, not the cryptographic PRNG algorithm itself.