Account lock after a single authentication failure

I have 2 standalone servers (not in any domain/AD) - Windows 2003 and Windows 2008.

When connecting to shares on them, using the UNC path from Windows 2000 pro, or Samba (under linux), if I make a mistake typing the password, these servers will lock the account on first attempt.

It works OK, when connecting from Windows 2003, Windows 2008 and Vista.

On all the connecting machines, it'll report auth failure, and will ask for credentials again. But from Windows 2000 and Samba it'll continue to fail on the second attempt even if the password is correct. Trying to log to these machines will report account locked.

The policy set is to lock account on 10 failures, and it behaves correctly when the connection attempt is done from W2008 or W2003 - i.e. even if the first attempt fails, the next one will succeed.

I've done the experiment to unlock the account, the from W2000 box I intentionally try to log with bad password, and even without second attempt I can see that the account is locked.

Any idea what I may change, so I avoid this problem? It usually happens when I'm in a hurry (of course :) ), so dealing with logging on the machines to unlock the account is really frustrating.

Solution 1:

I'm not sure what your level of familiarity is with Wireshark, so my apologies if this is too verbose or terse. ;-)

Essentially you'd want to install and start Wireshark on either end – it shouldn't matter which – and begin capturing on the appropriate interface:

1. In Wireshark, click the Capture menu, then Options…
2. Select the appropriate interface.
3. Since you mention you're not on a domain, I'd suggest entering a Capture Filter to limit the amount of captured traffic, something like:
   host 169.254.255.127
   The IP address should be for the remote host.
4. Start the capture, then attempt to browse the share.
5. Wait until the share appears, or you are presented with an invalid username/password box, then stop the capture.

I might suggest that you do this twice from each location; first, with a correct username and password; and second, with an invalid username and/or password. In between each attempt, be sure to disconnect from the remote share (so that it will re-request authentication), by doing:

net use \\remoteserver\SharedFolder /delete

What we are looking for here is SMB data, so the first way to prune out everything unrelated is to apply a simple Display filter in the Filter: field in the toolbars:

smb

In the beginning, you should see essentially three types of transmissions in the Info field:

• Negotiate Protocol Request/Response
  These show the protocols supported by the client, and which protocol was accepted by the server. You can expand this out in the Packet details section to see whether the server is using NTLM, Kerberos, etc.
• Session Setup AndX Request/Response
  This is authentication, which is hopefully happening in a challenge/response style. You'll see several possible Responses from the server:
  • STATUS_MORE_PROCESSING_REQUIRED: authentication is underway.
  • STATUS_LOGON_FAILURE: this is a logon failure – you will be most interested in these!
  • (nothing), in which case you can expand SMB > SMB Header and look for NT Status: STATUS_SUCCESS, which indicates a successful authentication.
• Tree Connect AndX Request/Response
  This is the client asking to connect to a particular location on the server; you'll typically see accesses to IPC$ as well as the share you really want.

So what you're actually looking for, then, are lines that say STATUS_LOGON_FAILURE. Then, look one line above that to see which user failed to authenticate.

Now, it's typical to see login failures whenever you browse a share; that's because Windows tries to authenticate using the logged-on user account first. So don't be surprised to see, even between the "modern" OS families (2k3, Vista, 2k8), logon failures.

When I ran this test earlier, I saw three logon failures with LOCALCOMPUTER\localuser before it even tried to use REMOTECOMPUTER\remoteuser (which, in my case, succeeded on the first try). And when I tried it against a machine on another domain, I got ten logon failures! That's before it even prompted me for alternate credentials.

If you want to filter out everything but authentication, change the display filter to:

smb.cmd == 0x73

To see a list of all logon failures, change the display filter to:

smb.nt_status == 0xc000006d

Keep in mind that these logon failures might not have any impact, as typically the account won't exist on the remote machine (so there is no "negative accounting", so to speak, to be done for it).

Edit: You might want to pause for a few seconds when it prompts for alternate credentials, so that you have a clear idea of which authentication failures belong to the "automated" attempts, and which belong to you deliberately mistyping the credentials.

Let us know what you find out!

Solution 2:

You are most likely having a problem with the difference between NTLM and NTLMv2. Windows 2000 and Samba default to NTLM which appears to a Windows 2003 and Windows 2008 server as a valid password attempt, but can't interpret the password correctly while expecting NTLMv2.

I believe that Windows 2000 can now support NTLMv2 along with the latest Samba bits...

You might try and configure your Windows 2000 and Samba to use NTLMv2...

Another strategy you might try is to downgrade the NTLM version on your Windows 2003/2008 servers to see if you still have the problems. I wouldn't leave it at NTLM, since NTLMv2 is much safer.