[cryptography] Kernel space vs userspace RNG

Krisztián Pintér pinterkr at gmail.com
Fri May 6 04:07:46 EDT 2016


On Thu, May 5, 2016 at 11:40 AM, shawn wilson <ag4ve.us at gmail.com> wrote:
> Just reflecting on the Linux RNG thread a bit ago, is there any technical
> reason to have RNG in kernel space?

i think it is worthwhile to separate the topic to entropy collection
and prng. prng is the easier part. the entropy collection is hard.
good arguments can be made to put the entropy collection into the
kernel, because

1, the kernel has access to much more entropy than any user mode program

2, user mode programs tend to have access to the same entropy,
therefore one program can recreate the prng state of another program
running on the same machine

as a sort of personal anecdote. out of fun, i once implemented a
program to create true random bitstream from sound card noise. to my
surprise, if i launch two instances of the program at the same time,
there is a good chance they will create the exact same output. it is
because the recording device serves the exact same audio data to all
programs doing the recording in parallel.

considering this, i would suggest rethinking the entropy estimates of
rngs, excluding everything a user mode program could possibly
eavesdrop on.


More information about the cryptography mailing list