GSSAPI security for connection encryption

[Douglas E. Engert]

Jeffrey Altman wrote:
> FTP AUTH GSS cannot be implemented with SEQUENCE protection if
> the client plans on sending data on the command channel while
> there is an active data channel using the GSS context.  This is
> a serious problem for GUI FTP implementations which issue
> STAT commands for status updates; or for other implementation
> wishing to protect against NAT/firewall timeouts.
> 
> I have a proposal to fix this but the changes are incompatible
> with the current implementation.  Hopefully I will find the
> time to work on submitting it to the IETF.

According to my sources, the Globus gridftp uses seperate contexts
for the control data channel, and uses SEQUENCE on both. Since FTP
is really a three party transfer, this make sense.


> 
> Jeffrey Altman
> 
> 
> 
> Markus Moeller wrote:
> 
> 
>>The MIT ftp client uses only the REPLAY flag, which I think should be
>>changed to the SEQUENCE FLAG to provide the correct protection for the data
>>channel.
>>
>>  gss_init_sec_context(&min_stat,
>>         GSS_C_NO_CREDENTIAL,
>>         &gcontext,
>>         target_name,
>>         (gss_OID_desc *)gss_trials[trial].mech_type,
>>         GSS_C_MUTUAL_FLAG | GSS_C_REPLAY_FLAG |
>>           (forward ? GSS_C_DELEG_FLAG :
>>     (unsigned) 0),
>>         0,
>>         &chan, /* channel bindings */
>>         token_ptr,
>>         NULL, /* ignore mech type */
>>         &send_tok,
>>         NULL, /* ignore ret_flags */
>>         NULL); /* ignore time_rec */
>>
>>
>>Will I get problems with the SEQUENCE FLAG if  I want to send a NOOP on the
>>command channel during a transfer happens on the data channel (.e.g. to keep
>>to command channel open through firewalls ) as the client and server have to
>>process the data in sync which is not necessarily given ?
>>
>>Thanks
>>Markus
>>
>>"Sam Hartman" <hartmans at MIT.EDU> wrote in message
>>news:tslzn4qssps.fsf at cz.mit.edu...
>>
>>
>>>>>>>>"Markus" == Markus Moeller <huaraz at moeller.plus.com> writes:
>>>
>>>   Markus> will Sequence protection (GSS_C_SEQUENCE_FLAG)cover replay
>>>   Markus> protection (GSS_C_REPLAY_FLAG) as well or are there cases
>>>   Markus> were I need both ?
>>>
>>>I'd recommend using both, but I believe sequence is typically a
>>>superset.
>>>
>>>________________________________________________
>>>Kerberos mailing list           Kerberos at mit.edu
>>>https://mailman.mit.edu/mailman/listinfo/kerberos
>>>
>>
>>
>>
>>
>>________________________________________________
>>Kerberos mailing list           Kerberos at mit.edu
>>https://mailman.mit.edu/mailman/listinfo/kerberos
>>
> 
> 

-- 

  Douglas E. Engert  <DEEngert at anl.gov>
  Argonne National Laboratory
  9700 South Cass Avenue
  Argonne, Illinois  60439
  (630) 252-5444