2.7.3 Handover Keying (hokey)

NOTE: This charter is a snapshot of the 73rd IETF Meeting in Minneapolis, MN USA. It may now be out-of-date.

Last Modified: 2008-03-04

Chair(s):

Charles Clancy <clancy@ltsnet.net>
Glen Zorn <gzorn@arubanetworks.com>

Security Area Director(s):

Tim Polk <tim.polk@nist.gov>
Pasi Eronen <pasi.eronen@nokia.com>

Security Area Advisor:

Tim Polk <tim.polk@nist.gov>

Mailing Lists:

General Discussion: hokey@ietf.org
To Subscribe: https://www.ietf.org/mailman/listinfo/hokey
Archive: http://www.ietf.org/mail-archive/web/hokey/current

Description of Working Group:

Most deployments of EAP in wireless networks employ an authenticator in
pass-through mode, usually located at the edge, coupled with a backend
AAA/EAP server.  Many EAP methods generate an MSK and an EMSK.  The
MSK is used by several EAP lower layers.  The EMSK remains at the peer
and server, but it is not presently used in any specifications. 
Different EAP lower layers make use of the MSK differently; the most
common usage is to derive Transient Session Keys (TSKs) to provide
access link security in networks (e.g., IEEE 802.11i, IEEE 802.16e),
although some lower layers (e.g., IKEv2) use the MSK for other
purposes.

Extensions to current EAP key framework will be needed to facilitate
inter-authenticator handover and roaming.  Some problems that need to
be addressed with extensions to EAP keying include:

1) Inter-authenticator handovers require re-execution of EAP
authentication even though the same EAP authentication server is
used.  Handover scenarios vary considerably in their fundamental
assumptions.  In scenarios where hosts remain connected during the
handover period, EAP authentication need not be in the critical path
for handover.  However, there are scenarios where necessary
connectivity is not available to support "make before break"
communications.
In these scenarios, significant handover latency can result.  To avoid
this latency, SDOs have employed methods such as context transfer and
anchoring that are inefficient or insecure or both.

2) EAP peers with unexpired keying material from a full EAP exchange
must take part in a full EAP exchange with the same AAA server to
extend a session. While some EAP methods provide fast re-
authentication mechanisms, a consistent, EAP-method-independent, low-
latency re-authentication mechanism is needed.

3) EAP generates keys (MSK and EMSK).  When the EAP WG updated the
protocol and specified the keying framework, many details regarding
the use of the EMSK were not specified.  The EMSK can be used as the
root of a cryptographic key hierarchy, and then the keys in the
hierarchy are used in various ways to provide the needed security
services.  In order to ensure that different keys derived from the
EMSK are cryptographically separate and that the key derivations are
coordinated in an acceptable manner, it is important to clearly
specify the top of the topology for the key hierarchy and some
guidelines for child key derivations.

4) When wireless networks employ AAA infrastructures, the cross-domain
roaming is handled by inter-domain authentication via the "home"
AAA/EAP server.  Any authentication must pass through the home server,
which increases latency. Latency can be reduced by establishing a
trust relationship between the EAP peer and the visited domain's
AAA/EAP server.  This trust relationship would be brokered by the home
EAP/AAA server.  Efficient re-authentication for the EAP peer can be
supported locally within the visited domain.

Some of the inconsistency in current handoff and roaming solutions can
be attributed to different trade-offs between computational cost,
mobility performance, and security.  Specifications are not consistent
in the way that the key derivation function (KDF) and KDF parameters
are used.  Clear direction by the IETF on the topology and
construction of a key hierarchy could reduce some of the
inconsistencies.  However, the HOKEY WG will not attempt to
standardize TSK derivation from the MSK, as this would interference
with work of other SDOs.

The solutions specified by the HOKEY WG fall into several categories,
based on timing and mechanism.  The authentication and key management
may occur before handoff, when latency is much less critical. 
Alternatively, authentication and key management can occur as part of
the handoff, where latency is critical.  Solutions should reduce or
eliminate the number of referrals to AAA servers, and solutions should
avoid re-executing lengthy EAP method exc hanges.This may be
accomplished by providing new mechanisms for cryptographic keying
material in combination with a protocol for the timely delivery of
appropriate keys to the appropriate entities.  Solutions are expected
to include "handover keying," "low-latency re-authentication,"
and "pre-authentication."

All solution categories are useful, each supporting different
scenarios.  The HOKEY WG may provide multiple solutions, each
addressing a different scenario.

Solutions specified by the HOKEY WG must:

1) Be responsive to handover and re-authentication latency performance
objectives within a mobile wireless access network.

2) Fulfill the requirements in draft-housley-aaa-key-mgmt and
draft-ietf-eap-keying.

3) Be independent of the access-technology.  Any key hierarchy
topology or protocol defined must be independent of EAP lower layers. 
The protocols may require additional support from the EAP lower layers
that use it.

4) Accommodate inter-technology heterogeneous handover and roaming.

5) No changes to EAP methods.  Any extensions defined to EAP must not
cause changes to existing EAP methods.

In specifying an access-technology-independent solution, media
independent guidelines for SDOs may also be needed to explain how the
keying material and signaling can be employed in a specific access
technology.


HOKEY WG Deliverables
=====================

All the specifications will be EAP-method-independent manner and
access-technology-agnostic.  EAP re-authentication and EAP pre-
authentication authenticator are expected to use the same layer and
the same protocol as the original EAP authentication used for the
authenticator.  They will provide enough semantics and guidance so
that all SDOs can employ them and preserve cryptographic separation.

1) A Problem Statement that defines the problem of re-authentication
and key management.  The discussion will include security and
performance goals for the solutions.  Too often, mobility optimization
discussions do not clearly identify the scenarios that are being
addressed; this lack of clarity often makes it difficult to agree on
whether the proposed optimizations offer real value.  To avoid this
situation, the Problem Statement must clearly describe the scenarios
that are being addressed, and the assumptions about the handover
environment associated with that scenario.

2) A specification of an EMSK-based key hierarchy topology.  The
specification will include a requirements, one or more KDF, and
parameters.  This is follow-on work EAP (RFC 3748) and EAP keying
framework that was developed in the EAP WG.

3) A specification for the derivation of root keys, such as the
handover root key (HRK), as well as any other media-independent keys
(e.g., authenticator level keys) that need to be derived from such a
root key, to support re-authentication and handover key management. 
The HOKEY WG can base these keys on either the MSK or the EMSK
produced by EAP (pick one).  If the consensus is to use the EMSK,
then the HRK forms one branch in the EMSK-based key hierarchy.  This
specification will describe the properties of each key that is
specified, and the topics must include caching, naming, scope,
binding, storage, and key lifetime. 

4) A protocol specification for media-independent, low-latency
re-authentication.  The protocol specification must support handovers
between EAP authenticators.  This protocol specification is expected
to employ a re-authentication branch in the key hierarchy.

5) A protocol specification for secure and timely distribution of
cryptographic keys to support roaming and handover.  Use of AAA
protocols is preferred, and if used, should leverage existing work if
possible (such as RADEXT WG work on RFC 3576bis and RADIUS
cryptographic algorithm agility). However, if AAA protocols cannot
meet the objectives, other protocols for reactive or proactive
distribution or retrieval of keys by the proper entities
is permitted.

6) A specification for inter-EAP-authenticator roaming and re-
authentication in visited domains that is brokered using inter-domain
trust relationships to support efficient inter-domain roaming.

7) A specification for EAP pre-authentication to support low-latency
inter-authenticator handoffs.

Goals and Milestones:

Done  First draft on EMSK-based Keying Hierarchy
Done  First draft with a problem statement on EAP re-authentication and key management
Done  First draft on EAP Re-authentication and Handover Keying Hierarchy
Done  First draft on EAP Re-authentication Protocol
Done  First draft on Protocol and Keying Hierarchy for Visited Domain Handovers and Re-authentication
Done  Submit EMSK-based Keying Hierarchy draft to IESG
Done  First draft on Handover Key Distribution Protocol
Done  Submit the problem statement draft to IESG
Done  Submit EAP Re-authentication and Handover Keying Hierarchy draft to IESG
Done  Submit EAP Re-authentication Protocol draft to IESG
Sep 2007  Submit Protocol and Keying Hierarchy for Visited Domain Handovers and Re-authentication draft to IESG
Done  First draft on EAP Pre-authentication Specification for inter-technology and inter-domain handoffs
Mar 2008  Submit EAP Pre-authentication Specification to IESG
Mar 2008  Re-charter or shut down WG

Internet-Drafts:

  • draft-ietf-hokey-reauth-ps-09.txt
  • draft-ietf-hokey-emsk-hierarchy-04.txt
  • draft-ietf-hokey-erx-13.txt
  • draft-ietf-hokey-key-mgm-03.txt
  • draft-ietf-hokey-preauth-ps-02.txt

    No Request For Comments

    Meeting Minutes


    Slides

    Key Management Presentation