Re: Flaw in "random32: update the net random state on interrupt and activity"

[Willy Tarreau <w@1wt.eu>]

Hi Ted,

On Wed, Aug 05, 2020 at 11:34:32AM -0400, tytso@mit.edu wrote:
> That being said, it certainly is a certificational / theoretical
> weakness, and if the bright boys and girls at Fort Meade did figure
> out a way to exploit this, they are very much unlikely to share it at
> an open Crypto conference.  So replacing LFSR-based PRnG with
> something stronger which didn't release any bits from the fast_pool
> would certainly be desireable, and I look forward to seeing what Willy
> has in mind.

I'll post a proposal patch shortly about this, hopefully this week-end
(got diverted by work lately :-)). Just to give you a few pointers,
it's a small modification of MSWS. It passes the Practrand test suite
on 256 GB of data with zero warning (something that Tausworthe is
supposed to fail at).

By default, MSWS *does* leak its internal state, as Amit showed us (and
seeing that the paper on it suggests it's safe as-is for crypto use is
a bit shocking), but once slightly adjusted, it doesn't reveal its state
anymore and that would constitute a much more future-proof solution for
quite some time. Tausworthe was created something like 20 years ago or
so, hence it's not surprizing that it's a bit dated by now, but if we
can upgrade once every 2 decades I guess it's not that bad.

Cheers,
Willy