Hi alain and every body
Concerning point I, let me go back to dvb-rcs : encryption at the
encapsulation layer
(which is now MPE but may become ULE in the future) has been provisioned
and the way it is
controlled by the 2 bits field in the MPE header has been clearly specified
(see my precedent mail)
I am not aware of the exact reasons which led poeple who contributed to the
dvb-rcs definition to do this
may be they did something wrong or expected a functionnality that will
never be used....!!!!
But it is more probable that they had real reasons to do this (so we can
avoid to reinvent the wheel as you said)
Any one has an idea?
I see an example which could make case 3 difficult to apply. It concerns
bridging
When forwarding (in the Gateway for example) is performed by bridging (as
in Ethernet switches) and not by routing (as routers),
IPsec can not apply because the GW handles frames and not datagrams (am I
wrong?). Such case may happen with e.g a PPPOE access scheme.
So if IPsec is not applied end-to-end (but this can not be controlled by
the satellite segment), traffic will be transmitted clear on the
sat interface
Concerning the use of the extension header, the main question is : when
encryption is applied at ULE level, do we need the encryption control field
in all SNDUs or not. If yes, the extension header solution may turn to be
inefficient since it would lead to about 5 bytes extra header (taking your
example below). I submitted this question to Alcatel's security experts but
would be great if others in this groupe could also think about it and make
suggestions
regards
Alain RITOUX <alain.ritoux@6wind.com>@erg.abdn.ac.uk on 26/02/2004 15:25:15
Veuillez répondre à ip-dvb@erg.abdn.ac.uk
Envoyé par : owner-ip-dvb@erg.abdn.ac.uk
Pour : IPDVB <ip-dvb@erg.abdn.ac.uk>
cc :
Objet : Re: Encryption control of SNDU
Gorry Fairhurst wrote:
So, I'm trying to build a list of "issues for ULE" and the questions I
have
are:
(i) Does the proposed ULE base header format (4/12B of header) need to be
changed to support any required encryption/scrambling?
Possible answers include:
* Yes - because the ULE header must not be increased when
link encryption is used.
* No - because the ULE header can specify a TYPE that could
indicate an encrypted payload, and hence this issue
could be solved by using an extension header of some form.
If it is YES, then this has design implications for the ULE Spec.
I would say No, because :
- I don't really understand the need for ULE level encryption
see point I)
- If needed, it can be separated from ULE base specs. see point II)
I) Why ULE level encryption ?
=============================
I still don't sse the need for an ULE-level encryption, because
I see 3 possibilities of encryption
1) At MPEG-2 level, i.e. same encryption for whole content of a PID
2) At ULE level, i.e. encryption on a per SNDU basis for the case as
Tarif said, where PID is mutli-receiver (belonging to different groups)
3) At IP level
but what can distinguish 2 SNDU ?
- IP addresses ?
- DVB Mac addr ?
But the DVB Mac addr is itself the result of IP address/prefix
resolution (which can use destination and/or sources)
So I think it can all be expresed in term of either destination and/or
source Addresses, which can perfectly be handled by IPsec. So 2) and 3)
seem to me to offer the same granularity and 3) is already defined and
working.
This is my current understanding, and a counter-example where 2) would
solve a use-case and not 3) would be very helpfull.
Anyway, I may have missed something, so :
II) Etxension Header for Encryption
====================================
*IF* an ULE encryption is needed, then some extension header can be
defined to provide it, so the SNDU would be like :
- [ULE_Header] [DVB_MAC*] [SEC_Header] [xxx SNDU Payload xxx]
|
+--> Type = sec_header
(*) if MAC addres bit is '0'
- The SEC_Header would be
+---+---+---+--// ..... // ----+
| NH | L | security params |
+---+---+---+--// ..... // ----+
with NH = Type of SNDU payload (IP, IPv6, ...)
L = Length of SEC_Header
security params will define key material, type of security
(encryption, authentication, ....), algo, whatever ...
In the extension headers I proposed, the SEC_Header would be one
of the "drop packets if unknown" type.
- The SNDU payload woul be clear/encrypt/padded accordingly to what
params will be found in SEC_Header.
With that approach, the only thing needed is to include the Extension
Headers mechanism definition in the ULE base specs. Then ALL that
SEC_Header stuff can be described in a separate doc, and good luck with
the (re)keying framework/protocols, security analysis ...
Regards.
Alain.
--
Alain RITOUX
Tel +33-1-39-30-92-32
Fax +33-1-39-30-92-11
visit our web http://www.6wind.com
T. Zein
ALCATEL SPACE
DRT/RST -- Ingénieur Systèmes
Tel : 0534356918 / Fax : 0534355560
Porte : W.220