Win 95 Problems

[Luke Kenneth Casson Leighton]

On Wed, 7 Jul 1999, Michael Stockman wrote:

> Hello,
> 
> The reason for the problems with samba, current HEAD, and W95 is due
> to an error in smbd/password.c. In function smb_password_ok() there is
> a piece of code:
> 
>   if (lp_server_ntlmv2() == False)
>   {
>     DEBUG(...);
>     return False;
>   }

yep, i know. i changed this to what is outlined below.  i added this check
yesterday as someone told me that they were using LANMAN1 with "server
ntlmv2 = true", which is a definite configuration error.

> The NT MD4 password check is done before this and the LM MD4 password
> check is performed after this.
> 
> What this code does is that if your samba is not configured as a ntlm2
> server (this is the default), LM MD4 passwords will not be checked.
> Since these are what W95 sends, it breaks. This should probably be
> removed or changed either to:
> 
>   if (lp_server_ntlmv2() == True)
> or
>   /* something related to the negotiated protocol,   */
>   /* I don't know enough about this to write a patch */
>   /* if this is the case */
> 
> This would allow older systems than ntlm2 to continue to use LM MD4
> passwords and (possibly, if the code is kept) prevent them in newer
> systems.
> 
> What I am unclear about is whether ntlm2 is actually a protocol level
> negotiated between the client and the server or something that we
> either are or aren't and everyone else will have to adjust to (which
> they won't?). Looking at the code today the latter seems to be the
> case, but I believe we will have huge integration issues between newer
> and older software if that is our path (so I hope I'm just
> misunderstanding something).

you're misunderstanding something.

ok, i was a little concerned about telling people about ntlmv2 (crypto
issues).  but actually, as it uses one-way hashes (a modified version of
hmac_md5, rfc2104.txt) there _are_ no crypto issues [if it used des or
rc4 (two-way, reversible systems) then that would be a different matter].

ntlmv2 is microsoft's latest, more secure authentication mechanism.  as it
used hmac_md5 twice, where md5 is 16 times more computationally expensive
than md4, it is more time consuming to brute-force.  the client and server
also mutually validate each other with this system.  the server-side
implementation in samba does not do any client validation, although i may
implement this at a later date [check the contents of the client
challenge].  this includes time validation, where the client and server
must be synchronised within a certain range (ms uses +/- 30 minutes).

the default behaviour is currently exactly the same as old versions of
samba: ntlmv2 is disabled by default in all client-side and server-side
code (rpcclient, smbclient, "security = domain", smbd).

MS KB article Q147706:

this article describes in detail how insecure LM#es are, and points you to
URLs where current information and statistics on DES cracking can be
obtained.  it outlines how to enable ntlmv2 and why this should be done.

key is:

HKLM\system\currentcontrolset\control\lsa\LmCompatibilityLevel (DWORD)

* enable ntlmv2 in nt clients by setting LmCompatibilityLevel registry
setting to 0x1 (equivalent to client ntlmv2 = auto), or 0x3 if you wish to
refuse to talk to down-level servers (equivalent to client ntlmv2 = true).

* enable ntlmv2 in nt servers by setting LmCompatibilityLevel registry
setting to 0x4 (equivalent to server ntlmv2 = auto), or 0x5 if you wish to
refuse to talk to down-level clients (equivalent to server ntlmv2 = true).

does this help?

luke

<a href="mailto:lkcl at samba.org"   > Luke Kenneth Casson Leighton    </a>
<a href="http://www.cb1.com/~lkcl"> Samba and Network Development   </a>
<a href="http://samba.org"        > Samba Web site                  </a>
<a href="http://www.iss.net"      > Internet Security Systems, Inc. </a>