[Cryptography] Duh, why aren't most embedded TRNGs designed this way?

[Peter Fairbrother]

On 23/05/2021 22:33, Robert Wilson via cryptography wrote:

But in the real world, wherever
> that is, we may sometimes have to accept [probabilistic proofs]. 


Indeed. I don't like proofs involving more than a simple infinity, but 
sometimes they have to be accepted too.

However in the case of TRNG's we have a requirement - unpredictability - 
which is not susceptible to statistical proof of any kind.

A quick proof of this (which I'll probably get wrong, I did say quick): 
Any statistical test can only test a finite stream. There are 2^N 
possible streams of finite length N bits, but there are an infinite 
number of predictable ways to generate each of the individual streams.


We can use statistical methods to show with some confidence whether or 
not a TRNG has a biased output - but we cannot use them to prove that it 
is unpredictable.

Von Neuman famously said "Anyone who attempts to generate random numbers 
by deterministic means is, of course, living in a state of sin."

The same might be applied to those who accept probabilistic proofs of a 
TRNGs unpredictability.



Peter Fairbrother