| |||||||||||||
|
| |||||||||||||
What's the problem?What's the problem?Posted Feb 14, 2003 17:58 UTC (Fri) by Ross (subscriber, #4065)Parent article: Mandrake security update to util-linux I thought that /dev/urandom was just as good as /dev/random unless someone has figured out how to break SHA-1? I'm not saying the patch was a good idea, but I'm not sure why reverting it couldn't wait until the next release.
(Log in to post comments)
What's the problem? Posted Feb 16, 2003 10:29 UTC (Sun) by Peter (guest, #1127) [Link] I thought that /dev/urandom was just as good as /dev/random unless someone has figured out how to break SHA-1? If you're not feeding new entropy into your pool, SHA-1 is just as predictable as any other PRNG. Breaking SHA-1 would be necessary if you wanted to re-derive the raw PRNG input, but that's not particularly useful in this context. I agree with you, though, that this doesn't appear to be very serious. In this case the random numbers are being used to produce a cookie for X11 authentication. Unless I'm missing something, this cookie can be sent in plaintext over the net - so the relative strength of /dev/urandom is most likely the least of your worries.
What's the problem? Posted Feb 16, 2003 23:43 UTC (Sun) by Ross (subscriber, #4065) [Link] I had read differently on the lkml a few years back. It was stated (IIRC) that /dev/urandom was safe on systems which had very little (or no) entropy gathered as long as:1) the initial entropy pool contents were unknown to the attacker 2) SHA-1 could not be reversed I don't see any flaws in the reasoning, but then again I'm not a cryptographer :) Pseudo-random number generators are often weak cryptographically, but the only inherent weakness is that knowing the internal state at any point in the past will allow you to predict the output at any point in the future (assuming no re-seeding is performed). Using /dev/random is probably better for generating keys out of paranoia. It is of course possible that someone will find a weakness in SHA.
What's the problem? Posted Feb 19, 2003 8:28 UTC (Wed) by Peter (guest, #1127) [Link] Well, on a scale of people who know anything about cryptographic PRNGs, I rate quite low. (: I had read differently on the lkml a few years back. It was stated (IIRC) that /dev/urandom was safe on systems which had very little (or no) entropy gathered, so long as: I don't see any flaws in the reasoning, but then again I'm not a cryptographer :) Ditto, squared. It sounds good anyway.
|
|
||||||||||||