Hi,

> I'm inclined to agree that IPSec closes more holes than ssh or ssl do. 
> However, I still see a couple of problems I haven't been able to see a
> fix for.  (Yes, this is a request for help!)
> 
> 1. Ideally, we could use IPSec to authenticate each *workstation* long
> before we ask for names or passwords.  However, each workstation starts
> up without a key every time it starts (it either builds one, or gets it
> from a central server somehow).  I think this means that we cannot use
> the IPSec keys to authenticate the workstation.  How can we prove that
> the workstation isn't an outsider trying to get into our network?

There is indeed a need for a key on a stable storage or a password protecting
the connection, anyway something which isn't fetched from the server.
 
> 2. How can we prevent man-in-the-middle attacks?  Is it possible without
> workstation authentication?

I think it is impossible to avoid man-in-the-middle attacks with tftp: the
tftp request is broadcast, and an attacker may have installed a malicious 
tftp server. If the attacker also increase the load on the real server such 
that it cannot respond to tftp requests, it should be easy to send a malicious
kernel/initrd to the workstation. At that point there is nothing that can be 
done to avoid the attacker doing anything he wants on the workstation with 
malicious kernel.

Pat


-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_____________________________________________________________________
Ltsp-discuss mailing list.   To un-subscribe, or change prefs, goto:
      https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help,   try #ltsp channel on irc.openprojects.net

Reply via email to