[OpenAFS-devel] openafs - proposed cache security improvement

[Marcus Watts]

> From: Jim Rees <rees@umich.edu>
> To: Marcus Watts <mdw@umich.edu>
> Cc: openafs-devel@openafs.org
> Message-ID: <20070330193421.GA8294@citi.umich.edu>
> References: <460BABA6.4080309@columbia.edu> <200703300255.VAA14052@quince.ifs.umich.edu>
> Date: Fri, 30 Mar 2007 14:34:21 -0500
> 
> The citi implementation of pkinit is in the MIT kerberos source tree, but I
> don't think it has made it in to an official release yet.  It has two
> interfaces for doing its pk work.  One is pkcs11, which can be used to talk
> to a smartcard or other secure hardware (or even software) token.  The other
> simply reads certs and keys out of a file.  It requires a client cert, not
> just a key.
> 
> Heimdal has its own pkinit implementation.  It interoperates with ours.  I
> don't know much more about it.
> 
> Microsoft of course has their own implementation which doesn't match the
> rfc.  We do, however, interoperate with them.  MacOS also has an
> implementation.  Last time I looked it was based on an early draft of the
> rfc but I'm sure that has changed.  It uses the Mac crypto api.
> 
> I don't think pkinit could be used to obtain a host context without a host
> key, but maybe someone could think of a way.

draft-ietf-krb-wg-anon-03.txt proposes a way a client could get a ticket
without a client side identity, using DH.  This actually looks more
useful than PKU2U.  However, it does require kdc changes; PKU2U merely
requires library changes + gssapi.

				-Marcus Watts