[28972] in Kerberos

home help back first fref pref prev next nref lref last post

Re: Query about an admin testing a user's creds

daemon@ATHENA.MIT.EDU (Christopher D. Clausen)
Mon Jan 7 00:03:23 2008

Message-ID: <79663E2A269A42EF819A317BD60E0E2F@CDCHOME>
From: "Christopher D. Clausen" <cclausen@acm.org>
To: "Coy Hile" <coy.hile@coyhile.com>, <kerberos@mit.edu>
Date: Sun, 6 Jan 2008 23:01:59 -0600
Cc: chile1@bloomberg.net, jmcvetta@bloomberg.net
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Errors-To: kerberos-bounces@mit.edu

Coy Hile <coy.hile@coyhile.com> wrote:
> If we need to test, for example, that a user is actually getting a
> TGT, we need to inform the user that we're changing their password
> temporarily, change it, authenticate as them directly, and then have
> them change it back.  We've all been wondering aloud whether there is
> some way for an admin to get creds as a user directly (Eg, something
> like su - user which actually does a kinit as that user).  Has
> something along those lines been implemented?  If not, what's the
> reasoning behind it not being so implemented? (I'm perfectly happy to
> accept "Because it's Really Stupid(tm) for the follwing reasons..." as
> an answer too :))

What flavor of Kerberos are you using?  I beleive that it is trivial 
with a Heimdal setup for a Kerberos admin to extract a keytab for any 
principal and NOT actually change the password of the principal.  (Use 
the ext_keytab command in kadmin.)  It is less easy with an MIT setup.

You can revert the krb5 database to the point it was at before a 
principal change, however if other principals were changed in the mean 
time, you could have a serious syncronization problem.  You may be able 
to do this manually by just finding the data in the dump for a 
particular principal and injecting it into a newer dump of the current 
Kerberos database.  I am unaware of potential fallout from doing this 
though.

Alternately, you could modify your change password procedure to either 
store the cleartext of the password (bad idea) or generate a keytab for 
the user using the provided password (slightly less bad of an idea) 
during the change process.

<<CDC 


________________________________________________
Kerberos mailing list           Kerberos@mit.edu
https://mailman.mit.edu/mailman/listinfo/kerberos

home help back first fref pref prev next nref lref last post