[Freedombox-discuss] Entropy from the soundcard
Daniel Kahn Gillmor
dkg at fifthhorseman.net
Fri Sep 16 13:25:03 UTC 2011
Frank from debian-nas.org wrote me off-list about entropy daemons,
and agreed i could share his notes publicly.
> haveged runs just fine on both my QNAP TS-109 (Marvell Kirkwood) and
> Linksys NSLU2 (ixp4xx) ARM computers running Debian Squeeze.
> The amount of entropy from havaged is at least one if not more orders
> of magnitude greater than what I get from a USB ADC with randomsound
> or timer entropy daemon (http://www.vanheusden.com/te/). I've tested
> the quality of the entropy with ENT (
> http://packages.debian.org/squeeze/ent), and it gives very good
> results (passes all its tests of randomness).
> Compared to haveged, randomsound and timer entropy daemon provide too
> little entropy to really make much of a difference on my ARM machines.
I asked him about what kind of auditing he'd done, (e.g. whether he'd
tried to bias the random pool through the processor for haveged), and he
> I haven't really looked into the details very much. All I did was measure the output speed. Here's the results:
> 335 kBps of entropy on a Marvell Kirkwood 1.2GHz ARMv5 CPU
> 76-80 kBps of entropy on a Linksys NSLU2 @ 266Mhz
> 1.0 kBps of entropy when using the cheap '3D SOUND' USB audio stick or when using the Hercules Muse Pocket USB as audio device on a Linksys NSLU2 @ 266Mhz
> Timer entropy daemon
> produces ~34 bps of entropy on a Linksys NSLU2 at 266Mhz, while using 13-18% cpu time
> produces ~62 bps of entropy on a Intel Xeon 2.4Ghz Xen VPS, using 5% cpu time
> As you can see, Timer entropy daemon produces an insignificant amount of entropy, Randomsound a bit more and haveged produces so much more that the other two are irrelevant.
Just passing on the interesting news,
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 1030 bytes
Desc: OpenPGP digital signature
More information about the Freedombox-discuss