Hi Ludwig, Thank you so much for your detailed review! We have addressed your comments in the -03 update submitted. It might be a bit hard to see from the diff, as we also made a restructuring based on comments at IETF105, but I answer to detailed comments inline.
Thanks again,
Francesca
On 19/07/2019, 13:21, "Ludwig Seitz" <ludwig.se...@ri.se> wrote:
Hello Francesca, Marco,
I have finally managed to read the whole of draft-ietf-ace-key-groupcomm
and have a few comments for you:
Figure 2: I suggest you move the "Defined in ..." to the left
they way it is now, it looks as if the Dispatcher was defined in the ACE
framework.
FP: Ok, done.
==
"3.1. MUST contain ... 'grant_type'"
This is no longer true. Grant_type was made optional in one of the
latest updates of the ACE framework. When it is absent,
grant_type=client_credentials is assumed as default.
FP: Ah, we had missed this, thanks for pointing that out! So fixed as you
suggest.
==
3.3 "... and includes the following parameters: "
How is that supposed to work? The framework defines sending of the token
to /authz-info as a CoAP POST, where the payload is the bytes of the
token. In order to include additional parameters you would have to
redefine this payload to be a CBOR map (as the OSCORE profile does).
FP: We do specify the content-format to be ace+cbor in that case, now we have
clarified that the payload is a CBOR Map, containing token plus those
parameters.
==
The whole section 3 talks about parameters send back and forth between
the client and the KDC without defining how these are carried. It seems
to be implied that there are CBOR maps in the payload, but that should
be made explicit, especially where it differs from what the framework
defines.
FP: Yes, we have now clarified that.
==
3.1 the text about the parameters in the client's post to the
/authz-info endpoint at the KDC talks about parameters "sign_info" and
"pub_key_enc" and claims they are specified in 3.3.1 and 3.3.2, but
these sections specify the parameters for the "AS request creation
hints" messages and not in this context. At least some clarification
should be added.
FP: Ok, we have now added some text about that at the end of sections 3.3.1 and
3.3.2.
==
Section 4
"If not previously established, the Client and the KDC MUST first
establish a pairwise secure communication channel using ACE."
This sentence is not strictly correct. Using what part of ACE? The ACE
framework just says that you should establish a secure communication
channel, it's the specific profiles that define how these channels are
established. Please add some clarification.
FP: Note that this text is now moved to section 4.2. We have clarified by
replacing with the 2 following sentences:
" If not previously established, the Client and the KDC MUST first
establish a pairwise secure communication channel (REQ15). This can
be achieved, for instance, by using a transport profile of ACE."
==
"The Client and the KDC MAY use that same secure channel to protect
further pairwise communications, that MUST be secured."
This is very questionable use of requirements language. How do I claim
or test conformance with the second MUST?
FP: Right. We have rephrased. The MUST was not supposed to be normative
" The Client and
the KDC MAY use that same secure channel to protect further pairwise
communications that must be secured."
==
Section 4:
"The same set of message ..." should be "messages"
FP: Ok, this text actually disappeared with the restructuring.
==
" Note that proof-of-possession to bind the access token to the Client
is performed by using the proof-of-possession key bound to the access
token for establishing secure communication between the Client and
the KDC."
This may or may not be true for a specific secure communication protocol
(e.g. think of DTLS with X.509 certificates without client
authentication). You need to require this from the underlying secure
communication protocol.
FP: Thanks for pointing this out. We replaced the previous text with the
following:
" The secure communication protocol is REQUIRED to establish the secure
channel by using the proof-of-possession key bound to the access
token. As a result, the proof-of-possession to bind the access token
to the Client is performed by using the proof-of-possession key bound
to the access token for establishing secure communication between the
Client and the KDC."
==
4.1 "The endpoint in the KDC is associated to the 'scope' value of the
Authorization Request/Response."
Associated how? This is too unspecific to lead to interoperable
implementations. I would like to see this association specified in detail.
FP: This text disappeared with the restructuring. Now we do give "default"
names for the endpoints at the KDC, similar to how ACE does it.
==
4.2 "as defined in the "ACE Groupcomm Key" registry, defined in Section
11.5."
If possible, rephrase this to avoid the double use of "defined".
FP: Ok, done.
==
5.2
"If the leaving node wants to be part of a group with fewer roles, it
does not need to communicate that to the KDC, and can simply stop acting
according to such roles."
There are legitimate cases where a node might want to explicitly
deactivate roles it is currently using (principle of least priviledge)
and not just stop using them.
FP: I will answer to this comment in a separate mail, to include Jim's point.
==
6. "Then, if it wants to continue participating in the group
communication, the node has to request new updated keying material to
the KDC."
should be "... keying material from the KDC."
FP: Ok, fixed.
==
Sections 8. and 9.
Would be nice if there also were back-references to where those
parameters are
defined in the draft.
FP: Now added.
/Ludwig
--
Ludwig Seitz, PhD
Security Lab, RISE
Phone +46(0)70-349 92 51
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Ace mailing list Ace@ietf.org https://www.ietf.org/mailman/listinfo/ace
