Title: The IETF Extensible Authentication Protocol (EAP) Working Group
1The IETFExtensible Authentication Protocol
(EAP)Working Group
- http//www.drizzle.com/EAP/int-03
- Interim Meeting
- Joint Discussion with IEEE 802.11i Ad Hoc
- October 15th, 2003
2Agenda, 0900-1200 (EAP WG)
- Preliminaries (10 min)
- Agenda bashing
- Blue sheets and notes takers
- Document status
- EAP Key Framework
- Open Issues (10 min) -- Bernard Aboba
- - http//www.drizzle.com/aboba/EAP/eapissues.htm
l - Conversation phases (20 min) -- Pasi Eronen
- Security Associations (50 min) -- Pasi Eronen
3IETF EAP WG andIEEE 802.11iJoint Discussion
- http//www.drizzle.com/EAP/int-03
- October 15th, 2003
4Agenda, 1330-1700 (EAP WG and 11i)
- Preliminaries (10 min)
- Agenda bashing
- Blue sheets, notes takers
- Document Status
- IETF Document Status and Process -- Bernard
Aboba - IEEE Document Status and Process -- Dave Halaz
- Specific Technical Issues
- Key scoping and fast handoff (30 min) -- Bernard
Aboba - Authorization issues and fast handoff (20 min)
-- Jari Arkko - Fast roaming implementation results (30 min) --
Bill Arbaugh - Open Discussion
- Conclusions and Next Steps
5Reading Materials
- Slides
- http//www.drizzle.com/aboba/EAP/int-03
- EAP Key Framework
- http//www.ietf.org/internet-drafts/draft-ietf-eap
-keying-00.txt - EAP State Machine
- http//www.ietf.org/internet-drafts/draft-ietf-eap
-statemachine-00.pdf - EAP Base Specification
- http//www.ietf.org/internet-drafts/draft-ietf-eap
-rfc2284bis-06.txt - EAP Open Issues
- http//www.drizzle.com/aboba/EAP/eapissues.html
- IEEE 802.11i Drafts
- Userid p8021, Password go_wildcats
- http//grouper.ieee.org/groups/802/11/Private/Draf
t_Standards/11i/802.11i-D6.0.pdf - http//www.ieee802.org/1/files/private/1-REV-draft
s/d7/802-1X-Rev-d7-1.pdf
6EAPStandards Development Status
7Dependencies
RFC 3575
RFC 3576
RFC 3579
EAP-SIM (and other methods)
1X-REV
EAP state machine
2284bis
RFC 3580
EAP keying framework
802.11i
RFC3588(Diameter base)
Diameter EAP
DiameterNASREQ
8Document Status Summary
- Published as an RFC
- RFC 3575
- RFC 3576
- RFC 3579
- RFC 3580
- RFC 3588
- In AD evaluation
- Draft-ietf-eap-rfc2284bis-06.txt
- In IETF Last Call
- Draft-ietf-aaa-diameter-nasreq-13.txt
- AAA WG work items
- Draft-ietf-diameter-eap-02.txt
- EAP WG work items
- Draft-ietf-eap-statemachine-00.pdf
- Draft-ietf-eap-keying-00.txt
9Authorization,EAP, and Fast Handoffs
- Jari Arkko
- Ericsson Research NomadicLab
- Bernard Aboba
- Microsoft
10Background
- Discussions about SSID and its role in
authentication and fast handoffs - Discussions about the semantics of
Session-Timeout and its role in fast handoffs - Current authorization processes in AAA protocols
- Definition of the PMKSA
- Various fast handoff proposals
- gt Issue 182 and Section 1.4 in keying-00.txt
11Classic Network Access Authorization
- Network access decision is not just about
authentication, it is about authorization as well - Typically, need to check that
- This user is a legitimate user for the this
network - The user has subscribed to the type of service
requested - What parameters, such as filters, the access
network needs when serving this user - The user is within time of day or concurrent
session limits - There are no fraud, credit limit, or other
reasons preventing access for this user - AAA makes these decisions
12Access Authorization Using AAA
- AAA makes the decision, but the process is
complex - The decision isnt necessarily done in a single
entity. Everyone on the path can affect the
decision. - We have both an on/off decision and parameter
determination - Decisions can be based on either static policies
or dynamic state - The decision criteria is not communicated to the
access network, only the decision itself - Currently there are no standardized attributes
for the criteria - Even the parameters might not be communicated
- If a request parameter is acceptable, there is no
need for this
13Fast Handoffs
- Definition Fast handoff network access where
EAP authentication and associated AAA passthrough
are bypassed - Depending on the adopted solution
- AAA key transport may also be bypassed in favor
of context transfer, or - a pre-emptive AAA key transport may occur
14Authorization Issues ...
- Fast handoffs raise several authorization issues
- Consistent application of session time limits
- When you move, your remaining session time should
not increase - Absolute vs. relative time
- Avoidance of privilege elevation
- Movement should not imply the possibility to e.g.
log on to a corporate network from a guest
network - Encoding of restrictions
- Need to know the used criteria
- Decisions based on dynamic state
- Network-wide state can typically only be taken in
account by the AAA server - Keeping the dynamic state valid
15Fast Handoff Correctness Criteria
- Definition A fast handoff mechanism is said to
be correct if it creates the same context as a
new AAA exchange would have - If this can not be reached, it may be desirable
to fail - If the new and old NAS different in their
capabilities, it is difficult to achieve
correctness - AAA servers often have conditional authorization
rules - RADIUS Access-Accept rules
- Known attribute for a service which is not
supported gt fail - Unknown attribute gt ignore
- These should apply even in a context transfer!
16Examples
- 1. NAS 1 with VLAN support -gt NAS 2 without
- Should not result in access outside the original
VLAN - 2. NAS 1 with confidentiality support -gt NAS 2
without - Should not result in sudden loss of
confidentiality - 3. Users subscription revoked at AAA, user
continues to get access via fast handoff - Chasing the user around and trying to disconnect
him
17Conclusions
- PMK SA is intimately tied to the authorization
decision by AAA - We can not bypass authorization issues when doing
fast handoff - Fast handoff most likely to succeed in homogenous
environment - Neither RADIUS or DIAMETER was designed for the
requirements that fast handoff sets for them - Still, we need correct fast handoff mechanisms
- Allow-Fast-Handoff attribute from AAA is one easy
solution - Would avoid many of the above problems
- Only sent when dynamic state etc not an issue
- Disallow-Fast-Handoff is another solution
- Problematic if dynamic state or decisions
involved and DFH not set - But deployment of FH is easier even if home AAA
does not care about it - In addition, we may need
- Authorization attributes
- Authorization criteria attributes
- Get the details of context transfer acceptance
decision right
18Action Items
- EAP keying draft and 11i draft should clarify
that authorization parameters are a part of the
PMKSA - The definition of the AAA-Token should include
authorization parameters in Diameter EAP - Existing authorization parameters listed
- New ones that we need should be defined
- Authorized-Calling-Station-Id
-
- Disallow/Allow-Fast-Handoff attribute is needed
(but when?) - Fast handoff designers should take these issues
into account
19Backup Slides for Other Issues
- Jari Arkko
- Ericsson Research NomadicLab
20Diameter EAP Issues
- Basic EAP processing rules as in RFC 3579
- Added Redirect support to avoid revealing session
keys to nodes that have no need to know them - Authorization of AAA nodes to act as
authenticators or backend servers - Problem comes up in particular with Redirect
- Several possible technical solutions exist,
including local configuration and certificate
extensions - Proposed way forward is to document the issue but
not mandate a particular solution - AAA-Token contents
- The raw key (construction formula defined by
eap-keying) - Authorization AVPs
21EAP Keying Document Editorial Issues
- Reordering the document
- Starts with the solution and ends with
requirements - But hard to move sections since they have
dependencies - IEEE specific issues
- Necessary to keep examples
- But some text might be better in an appendix
- Suggestion PMK definition and IEEE terms only in
an appendix, the rest of the examples would stay?
22EAP Keying Document Small Technical Issues
- Are we really disallowing link layer specific
identifiers - Section 2.1.1 says so
- Even in the generic fashion as is done in EAP
Archie? - Multicast SA text needs work
- Are these derived from unicast SAs or PMKSAs?
- Multicast SAs contain common material for several
EAP authenticated clients and PMKSAs - No IANA considerations?
- What about PRF(ltsomethinggt,Constant)?
23Backwards Compatibility and Scoping
- Backwards compatibility issues need to be
addressed - We currently use the plain key in AAA-Key
- We (may) now switch to a scoped key calculated
via a hash - These are not interchangeable
- Client cant use hash(key) if server uses key
- Need to be signaled somehow for the client and
the AAA server - AAA server needs to know if AP is doing WPA or
final 11i - Cant wait until the 4-way handshake to find out
-- use beacon frame - If someone gets the plain key, he can manufacture
scoped keys - The use EMSK as a starting point for the scoped
key would prevent this - Cases
- Use the old key through an old access point
- A new access point can upgrade old AAA servers
key - Bidding down issues involved
- But RSN IE is already protected by WPA, may help
here...
24Parameter Agreements and Scoping
- Agreement over (some) parameters may be important
- SSID is currently the only really significant one
- Service type might be important, if more than one
service allows fast handoffs - MAC addresses might be important
- Do we need this?
- Two ways to achieve secure binding to the
parameters - 1) Mix the parameter information to AAA-Key
(PMKSA key) - 2) Securely exchange the parameters over EAP,
either using a generic approach (TLV/EAPv2) or
method-specific mechanisms (Archie) - Implications of the choice
- With mixing, new parameter introduction is hard
- Secure EAP exchange of the parameters does not
exist at the moment - But, we could add it to SIM, AKA, TLS, and
Archie -- would that be enough?
25SA definition changes in eap-keying
- EAP SA
- AAA-Key SA
- Unicast Secure Association SA
- Multicast Secure Association SA
- gt
- Long-term EAP SA
- Service SA ( PMKSA)
- Service Sub-SAs ( PTKSA, GTKSA)
- (AAA-Key is an algorithm and a protocol, not an
SA)
26SA definition changes in 11i D6
- MAC addresses away from PMKSA
- Clarify authorization parameter role in PMKSA
- Need references between PMKSA - PTKSA/GTKSA for
re-authentication and SA removal purposes - May need an accounting context reference in PMKSA
27IEEE EAP Issues Mentioned Tuesday
- Rejected in 6.1
- Use EAP Name for PMKID
- Remaining in 6.1
- Do not refer to EAP keying doc so EAP security
association is not defined - PMK naming / EAP naming. Current PMKID problem
for PSK - 8.4 EAP security association