[Cryptography] Is Ron right on randomness

[Ray Dillinger]

On 11/26/2016 04:30 PM, Ron Garret wrote:

> 3.  If OpenSSL wanted to include support for an external hardware source of randomness (like, say, the SC4-HSM) I wouldn’t mind ;-)

The best use for an HWRNG, in my opinion, is with a driver that writes
its output at some "reasonable" rate directly to /dev/random.  OpenSSL
would be supported automatically because it normally reads that device.
But that's just my opinion.  Others may have a different opinion because
that would require the /dev/random mixer to be run on all those bits,
which on some resource-constrained platforms might raise objections.

If you're serious about OpenSSL having direct hardware-generator support
to address those objections, I suggest you implement this yourself and
submit it as a patch.

Your self-interest in it is completely obvious, of course, but doesn't
represent a conflict of interest as such; there is no valid reason for
OpenSSL to NOT support hardware randomness sources, and by virtue of
being there first, any free code you submit could well define a common
interface to "standard OpenSSL-compatible" hardware random sources.

A standard interface for hardware bit generators is badly needed;
offerings from different vendors being incompatible is an unnecessary
source of difficulties and requiring custom code to be installed for
each is an unnecessary source of trust issues.

To address legal trust issues, it will at the very least be necessary
for you to make it clear that anyone is welcome to build compatible
devices and may do so without incurring liability for any kind of
intellectual property claim.

To address crypto trust issues, don't freeze out other sources. Write a
patch that would use the HWRNG as one of multiple bit sources rather
than as the exclusive source.

				Bear

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <http://www.metzdowd.com/pipermail/cryptography/attachments/20161126/106f1155/attachment.sig>