Re: [Emu] Consensus call on EAP-TLS key derivation

[John Mattsson <[e-mail]>]

I don’t see any strong reasons to keep the -15 key derivation. I started to prepare a PR for the likely change back to -13.

https://github.com/emu-wg/draft-ietf-emu-eap-tls13/pull/68

- Version 15 has the following wrong text that need to change. Key_Material can now be kept, but IV should be removed.

  “the Key_Material, IV, and Method-Id SHALL be derived”
  ”derivation of Key_Material, IV and Method-Id”

- The IANA table need to change.

- I would suggest to have the text agreed in -13 stating stating that the derivation from Key_Material is the same as in RFC 5216.

       The MSK and EMSK are derived in the same manner as with EAP-TLS [RFC5216], Section 2.3. The definitions are repeated below for simplicity:

       MSK = Key_Material(0, 63)
       EMSK = Key_Material(64, 127)

John

From: Emu <[e-mail]> on behalf of Joseph Salowey <[e-mail]>
Date: Sunday, 9 May 2021 at 19:54
To: EMU WG <[e-mail]>
Subject: [Emu] Consensus call on EAP-TLS key derivation

We had discussion on the list on whether to include context in the key derivation, but we never closed on the issue of separating out the MSK and EMSK derivation.  As a result several implementers have gone down the path of implementing what is in draft 13 and not separating out the derivation.  The main difference is that draft 15 separated out the EMSK and MSK derivation using two different labels while draft 13 used a single label to derive key material which is partitioned into two keys.   The reason for the change was to enable different access control for these two different quantities for different callers, however in practice it is EAP-TLS application which needs access to both keys that is the caller of the TLS library so this separation is not particularly useful.   Therefore the recommendation is to align with implementation and derive the MSK and EMSK by partitioning the key material from the key material produced by a single label of the exporter function.

Please respond to the list if you support the change below or not to revert some of the text in the key derivation section.  If you object to the change please state why.  Please respond by May 20,2021.

Thanks,

Joe

The proposal is to use the following key derivation which is largely a reversion to draft 13:

Draft-15 Text:

Type-Code = 0x0D

MSK        = TLS-Exporter("EXPORTER_EAP_TLS_MSK",Type-Code,64)

EMSK       = TLS-Exporter("EXPORTER_EAP_TLS_EMSK",Type-Code,64)

Method-Id  = TLS-Exporter("EXPORTER_EAP_TLS_Method-Id",Type-Code,64)

Session-Id = Type-Code || Method-Id

Proposed New Text:

Type-Code = 0x0D

Key_Material = TLS-Exporter("EXPORTER_EAP_TLS_Key_Material",

                               Type-Code, 128)

MSK          = Key_Material(0, 63)

EMSK         = Key_Material(64, 127)

Method-Id    = TLS-Exporter("EXPORTER_EAP_TLS_Method-Id",

                               Type-Code, 64)

Session-Id = Type-Code || Method-Id



The rest of the text of the section remains the same as draft-15.