Re: Bugs? in Microsoft RDP protocol, & Questions.

From: Patrick Chambet (patrick.chambetat_private)
Date: Fri Jan 18 2002 - 06:43:37 PST

  • Next message: Tekno pHReak: "Timbuktu DoS vulnerabilty"

    The data sent over the network doesn't seem to depend on the security
    level you have configured on your Terminal Server: the data is sent
    before the encrypted phase begins (before the session key negotiation).
    
    You can see in clear text the client name and the server license ID root
    (52310-005-2479922-00001 instead of 52310-005-2479922-04749 for
    example), but also the server domain, the server name and the server IP
    address after the "ncacn_np:" named pipe keyword:
    
    ncacn_np:194.41.26.111
    
    You can also observe some data that look like a public key exchange.
    
    For more information about exchanged data, you can try to get the
    Microsoft RDP specification document. This document isn't public and I
    don't have it yet: see
    http://www.microsoft.com/windows2000/server/evaluation/news/bulletins/rd
    pspec.asp
    
    Or you can look at the rdesktop source code:
    http://www.rdesktop.org
    
    ___________________________________________
    Patrick Chambet - MCP
    IT Security Consulting
    EdelWeb - ON-X Consulting Group
    http://www.edelweb.fr - http://www.on-x.com
    
    
    ----- Original Message -----
    From: "Pybus, David" <DPybus@colt-telecom.com>
    
    
    What security level have you set the terminal server to as if it is set
    to
    low it will be sending back a lot more than just its machine name
    unencrypted?
    
    Normally you wouldn't expose Terminal Services to the net so exposing
    things
    like a machine name are no worse than in the NetBios situation you
    mentioned. More importantly when exposing a TS machine to the net by
    default
    you give anyone who can connect the opportunity to brute force the local
    administrator account. This has to be prevent by configuring Terminal
    Services not allow the local admin account to logon and using other
    accounts
    instead which can be configure to lock after three failed attempt, or
    whatever else your policy dictates.
    
    Also something I have never seen anything about anywhere is how Terminal
    Services implements its key generation/exchange. As there is no
    indication
    that any type asymetric authentication occurs it seems reasonable to
    assume
    that Terminal Services are also probably vulnerable to man in the middle
    attacks.
    
    Food for thought,
    David Pybus
    



    This archive was generated by hypermail 2b30 : Sat Jan 19 2002 - 11:08:49 PST