Openwall GNU/*/Linux - a small security-enhanced Linux distro for servers
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Mon, 20 Jun 2011 19:19:13 +0400
From: Solar Designer <>
Subject: Re: CVE request: crypt_blowfish 8-bit character mishandling

On Mon, Jun 20, 2011 at 02:56:28PM +0000, The Fungi wrote:
> Would it make sense to include transitional compatability calls
> which preserve the original behavior?

Maybe, but this sounds worse than my "$2x$" proposal, which allows for
the same and more (it also lets one access the backwards compat
functionality without patching any code, by patching the hash encodings
in a database instead).

If an app knows what it is doing (and you're talking solely about such
apps above), it can simply replace 'a' with 'x' before its call to
crypt() or the like.

> Then applications using the
> library can be adjusted to fall back on the buggy version if the
> supplied data has 8-bit characters and the corrected calls don't
> result in a match.

This doubles the CPU time that a DoS attacker can consume per
authentication attempt, thereby halving the maximum iteration count that
an admin can reasonably set for new password hashes to use.  On the
other hand, if the iteration count was set significantly below the
affordable maximum for whatever reason, which is the common case, then
this is acceptable.  Then there's also the remotely measurable timing
difference, but that leak may be acceptable (telling an observer roughly
what goes on).

> This would allow tools to regenerate and replace
> non-conforming hashes if they were the result of this bug, and might
> make it easier to audit existing lists for them as well.

This is possible with my proposal as well.  The difference is that with
your proposal all hashes would remain listed as "$2a$", and only the
affected ones would be replaced (still remaining at "$2a$", which is
important not to leak any extra info about the passwords via the hash
encodings).  However, if one wants to implement your approach (or
similar), they can do so via my proposed interface (with "$2x$") as
well, by changing the 'a' to 'x' for just one function call (rather than
in the database).  The choice is theirs.

That said, I appreciate you posting this suggestion, and I'd be happy to
consider some more.  It is always possible that there's some brilliant
idea I had not thought of...



Powered by blists - more mailing lists

Your e-mail address:

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Powered by Openwall GNU/*/Linux - Powered by OpenVZ