Cannot authenticate with client keytab and AES128/256 against Active Directory

[Osipov, Michael]

> On 07/29/2015 07:43 AM, Osipov, Michael wrote:
> > add_entry -password -p osipovmi at COMAPNY.NET -k 1 -e
> > aes256-cts-hmac-sha1-96 add_entry -password -p osipovmi at COMAPNY.NET -k
> > 1 -e aes128-cts-hmac-sha1-96 add_entry -password -p
> > osipovmi at COMAPNY.NET -k 1 -e arcfour-hmac
> [...]
> > kinit: Invalid argument while getting initial credentials
> 
> Your primary problem here has to do with salts.  From the trace logs you provided
> me, the salt string for this principal was constructed using the principal name
> michael.osipov at COMAPNY.NET (not the actual realm name), not
> osipovmi at COMAPNY.NET.  ktutil unfortunately has no way to specify the salt
> string or to retrieve it from the KDC; it can only use the default salt for the principal
> name when adding a keytab entry using a password.  The RC4 enctype does not
> use the salt, so you don't encounter this problem when using only an RC4 key.

I am afraid you are right. Surprisingly, I read MS-KILE, chapter "3.1.1.2 Cryptographic Material"
and it does build the salt just like MIT Kerberos does. I see currently no reason why this happens
to my account.

I need to mention that we have a company-wide UPN suffix and every employee has an
enterprise principal in the form of <firstname>.<lastname>@company.com (same as email
address).

Moreover, I will try a few other accounts and will give you notice.

Unfortunately, I have no idea how Windows obtains the "custom" salt to derive the password
but would it be feasible to modify ktutil to receive a custom salt?

> I believe that people generally have better luck with msktutil for creating keytabs
> for use with Active Directory; it may solve this problem.

I will try that as soon as I get it compiled on FreeBSD. The port does not compile.
I was able to compile that with code modifications but unfortunately, this tool does not do
the same as ktutil. I cannot simply create a keytab for a user account. It constantly tries
to manipulate the machine account or create a computer/service account.


> The secondary problem is that you are getting the error message "Invalid
> argument" instead of something more accurate, like "Password incorrect"
> or "Preauthentication failed."  I don't know the exact cause of this problem yet,
> though I believe it has to do with our PKINIT code.

I guess this can be improved, can't it?

Michael