Jorge, Joe, thank you for your comments. So the updated proposal should be:
1) In Section "4.2.13. Crypto-Binding TLV" make "EMSK Compound MAC" and
"MSK Compound MAC" fields variable length depending on the number of output
bits of the MAC function negotiated by TLS protocol. A field "Compound MAC
Length" of size 1 octet is added after the "Nonce" field to represent the
length of each of "EMSK Compound MAC" and "MSK Compound MAC" fields. The
minimum length for each Compound MAC field is 20 octets (to honor the
original RFC fixed size and to allow usage of SHA-1 - that is still used in
TLS 1.2 - to provide MAC without padding). The length should never be less
than 20 octets since there are no ciphers supported by TLS 1.2 with a
digest of less than 160 bits output size. Therefore if "Compound MAC
Length" field value is less than 20 then the Crypto-binding TLV should be
considered of erroneous format and handled as described in section 3.6.
Error Handling. The value of the length field of the whole Crypto-Binding
TLV should be calculated accordingly (currently fix 76 octets).
The updated structure of Crypto-Binding TLV:
The Crypto-Binding TLV is defined as follows:
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|M|R| TLV Type | Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Reserved | Version | Received Ver.| Flags|Sub-Type|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ Nonce ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Comp MAC Len |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ EMSK Compound MAC ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| |
~ MSK Compound MAC ~
| |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
If we would not bump the TEAP protocol version - we still can change
the Crypto-Binding TLV version to 2.
2) [This was accepted and stays unchanged]
3) In Section "5.3. Computing the Compound MAC" when specifying the
list of fields to be placed in the BUFFER" should say "...2 A single
octet contains TEAP EAP method type 0x37". However to let this byte
play a role in protection of Crypto-Binding TLV I would suggest to
place in it the inner EAP method type related to the calculation or
0x0 if no inner EAP method was executed. What are your opinions on
this?
Thanks
Oleg
On Tue, May 26, 2020 at 12:29 PM Jorge Vergara <jover...@microsoft.com>
wrote:
>
>
> *From:* Joseph Salowey <j...@salowey.net>
> *Sent:* Monday, May 25, 2020 9:27 PM
> *To:* Jorge Vergara <jover...@microsoft.com>
> *Cc:* Oleg Pekar <oleg.pekar.2...@gmail.com>; Jouni Malinen <j...@w1.fi>;
> EMU WG <emu@ietf.org>
> *Subject:* Re: [Emu] TEAP - RFC 7170 - Errata ID 5768
>
>
>
>
>
>
>
> On Fri, May 22, 2020 at 1:18 PM Jorge Vergara <jovergar=
> 40microsoft....@dmarc.ietf.org> wrote:
>
> My review of this errata and the current responses from Oleg:
>
>
>
> 1. I agree with this proposed resolution. I do think this is an
> important omission that needs to be clarified in the RFC. Otherwise it is
> somewhat guesswork that truncation is the right action. I think the current
> wording leans toward truncation, but I definitely asked this question
> myself while implementing.
>
> [Joe] Why not just change the TLV to be variable length? It seems if we
> hardcode the length to 100 we risk having the same problem in the future?
>
>
>
> [Jorge] I have to admit that in my original response my brain skipped over
> the TLV length adjustment and focused only on the truncation. If the TLV
> length is going to be adjusted at all, I am in favor of making the TLV
> variable length. A minimum length could also be added (32 bits seems to be
> the current recommendation). However, I do believe this is a breaking
> change to any TEAP implementation – would the TEAP version need to be
> bumped?
>
>
> 1.
> 2. This bleeds into Alan’s TLS 1.3 document somewhat, but I agree with
> Jouni that this will need to change when the rest of the document is
> eventually updated to TLS 1.3.. There are enough TLS 1.3 related things to
> address in TEAP that I don’t exactly view this as an errata. I view it as a
> needed update, whether in this document, Alan’s document, or both.
>
> [Joe] I tend to agree that this is not an errata. However an update to
> TEAP should address these.
>
>
>
>
> 1.
> 2. Agree with Jouni that I don’t see the point of the 0x37 octet, but
> regardless this clarification of how it is encoded is positive (minor)
> change.
>
> [Joe] I think the original reason to include the TEAP method ID in the
> specification was to make sure that we differentiated between similar
> crypto binding implementations in other protocols such as EAP-FAST. I
> don't think there is much ambiguity here, but I am OK with including 0x37
> in the description.
>
>
>
>
>
> Thanks,
>
> Jorge
>
>
>
> *From:* Emu <emu-boun...@ietf.org> *On Behalf Of *Oleg Pekar
> *Sent:* Tuesday, May 5, 2020 6:27 AM
> *To:* Jouni Malinen <j...@w1.fi>; EMU WG <emu@ietf.org>
> *Subject:* [Emu] TEAP - RFC 7170 - Errata ID 5768
>
>
>
> Hi Jouni,
>
> I propose the following fix for the issues described in this errata id:
>
> 1) In Section "4.2.13. Crypto-Binding TLV" make "EMSK Compound MAC" and
> "MSK Compound MAC" fields 32 octets long (currently 20 octets). The MAC
> value is truncated at 32 octets if it is longer than 32 octets or padded to
> a length of 32 octets with zeros to the right if it is less than 32 octets.
> The length of the TLV should be changed to 100 bytes (currently 76).
>
>
>
> The motivation is to keep collision-resistance strength of MAC on 128 bit..
> Hash value truncation is described in "NIST Special Publication 800-107
> Revision 1: Recommendation for Applications Using Approved Hash Algorithms"
> <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fnvlpubs.nist.gov%2Fnistpubs%2FLegacy%2FSP%2Fnistspecialpublication800-107r1.pdf&data=02%7C01%7Cjovergar%40microsoft.com%7C6102b418ffd64be20f0b08d8012d1ade%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637260640995993815&sdata=MIOfT9tSEHTmbuP2oTjAxlgJdBOke3V4uAqQq7YDzN0%3D&reserved=0>
>
>
>
>
> 2) In Section "5.3. Computing the Compound MAC" specify that "MAC is the
> MAC function negotiated in TLS of TEAP Phase 1" (currently it says TLS
> 1.2)
>
>
>
> The motivation is to support TLS 1.2, 1.3 and possibly later TLS versions..
>
>
>
> 3) In Section "5.3. Computing the Compound MAC" when specifying the list
> of field to be placed in the BUFFER" should say "...2 A single octet
> contains TEAP EAP method type 0x37". Alternatively it could be "...2 A
> single octet contains EAP Type of the inner EAP method related to the
> calculation or 0 if no inner EAP method was executed" (currently "...2 The
> EAP Type sent by the other party in the first TEAP message")
>
>
>
> Please note that there's still a discussion on sending Crypto-Binding TLV
> on "Authentication inner EAP method" or "Inner EAP method that exports MSK"
> only.
>
>
>
> Thanks
>
> Oleg
>
> _______________________________________________
> Emu mailing list
> Emu@ietf.org
> https://www.ietf.org/mailman/listinfo/emu
> <https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.ietf.org%2Fmailman%2Flistinfo%2Femu&data=02%7C01%7Cjovergar%40microsoft.com%7C6102b418ffd64be20f0b08d8012d1ade%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637260640996003807&sdata=kBRkWM2GADS2iZP%2F6A9Ygr61JO6hcEe8cNK2Z6p9Pg0%3D&reserved=0>
>
>
_______________________________________________
Emu mailing list
Emu@ietf.org
https://www.ietf.org/mailman/listinfo/emu
