Subject: Re: Question on iSCSI security
To: "Williams, Jim" <Jim.Williams@Emulex.com>
Cc: "'ips@ece.cmu.edu'" <ips@ece.cmu.edu>
Message-ID: <OF9E1A4EEE.F624DBE4-ONC2256D43.00720CD6@telaviv.ibm.com>
From: "Ofer Biran" <BIRAN@il.ibm.com>
Date: Fri, 13 Jun 2003 00:30:54 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=us-ascii
Sender: owner-ips@ece.cmu.edu
Precedence: bulk

This is not be possible if A uses different secret/ticket
for B and C in the in-band authentication (for Kerberos ticket
it's always different).  Also it's not a classic man-in-middle since
A attempts to log into B (and authenticates B on the IPsec level)
and not into C, so this attack can actually  happen only after a previous
attack of compromising a legitimate server B.

  Regards,
    Ofer


Ofer Biran
Storage and Systems Technology
IBM Research Lab in Haifa
biran@il.ibm.com  972-4-8296253



                                                                                                              
                      "Williams, Jim"                                                                         
                      <Jim.Williams@Emu        To:       "'ips@ece.cmu.edu'" <ips@ece.cmu.edu>                
                      lex.com>                 cc:                                                            
                      Sent by:                 Subject:  Question on iSCSI security                           
                      owner-ips@ece.cmu                                                                       
                      .edu                                                                                    
                                                                                                              
                                                                                                              
                      12/06/03 21:50                                                                          
                                                                                                              
                                                                                                              





I am not up to speed on security and IPSec, so
there is probably a simple answer to this.  I
would be curious to know what it is.


Scenario:

A is an unwitting initiator, B is a malicious
target, and C is a victim target.

A attempts to log into B using IPSec.  B establishes
IPSec SA with C.  B is honest to IKE about its identity.
After establishing SA, B attempts to log into C, but
lies to the iSCSI layer and claims to be A.
B uses classic man-in-the-middle technique to get
A to respond to C's login challenge.  If this
works, then B has successfully logged into C
as A.

There are a number of similar scenarios with the
common thread that the attacker is truthful about
his identity to the IPSec layer, but lies about
his identity to the iSCSI layer.

These attacks are easily defeated if the iSCSI
layer cross checks remote end's identity with the
IPSec layer.  But it is not clear how this is done
and whether it will be done or is required to
be done.

If the IPSec layer verifies that the IP address
INSIDE the tunnel really belongs to B, and iSCSI
verifies that the IP address it sees really belongs
to A, and the data consulted for the verification
is secure, then one of these checks should fail,
but this seems like a stretch.

But perhaps I am missing something simple.