[onerng talk] Chaoskey presentation at LCA2017

[Paul Campbell]

(sorry I meant to respond earlier)

On Friday, 16 June 2017 4:57:38 AM NZST Peter Gutmann wrote:
> If DieHarder says it's random, it's probably random.

I think the main thing to realise is that nothing is ever really perfectly 
random - and learn how to live with that - in particular one's  goal should be 
to characterise  the lower bound on its entropy and collect enough extra 
entropy to live with that ....

In the case of OneRNG random-tools allows us to specify how much entropy (in 
fractions of a bit) we are feeding the kernel, they use this number to figure 
out how much extra data to fetch from your random source to fill /dev/random's 
needs .... in the latest OneRNG software release (from a few months back) I 
added a way to tune this: if you want to tweak this rate (maybe you're being a 
little extra paranoid, or maybe just you're OneRNG isn't being used much) you 
can edit /etc/onerng.conf and change:

	ONERNG_ENTROPY=".93750"

and change it to something like:

	ONERNG_ENTROPY=".5"

The ".93750" is just the value we got from testing, you probably shouldn't 
make it larger than that (it must be > 0 and < 1) but making it smaller will  
increase the amount of data that's pulled from OneRNG and stirred into the 
kernel's entropy pool

In short there are no perfect entropy sources, if you're worried you can 
always stir in more entropy than you need (assuming you have it)

	Paul