Federated Wireless
Network Authentication
Kevin Miller • Duke University
kevin.miller@duke.edu
Internet2 Joint Techs
Salt Lake City
February, 2005

Vision
Enable members of one institution to authenticate to the wireless network at another institution using their home credentials.
Often called the “roaming scholar” problem in HiEd.
Wired networks handled as well.

Framing the Solution
802.1x
Often used with WPA or WPA2 (802.11i)
Or middlebox access controller
EAP authentication
Exact EAP type selected by home institution, deployed on client machines
Phase 1: “Simple” RADIUS peering
Integration with existing authn backend

Topics
Federations
Wireless Security
802.1x
Working Group Activities
Project Plan: Phase 1, 2
Timeline
Deliverables
Administrivia

Federations
Goals of federations
Establish trust between entities
Make assertions about identities (authenticate) and release attributes
Protect user privacy through opaque user handles and controlled attribute release

Federations
All are relevant to FWNA
Want to leverage federation trust mechanisms instead of sharing RADIUS keys
Visited sites may want attributes about visiting users (e.g. type of user, mobile number)
Control release of identifiable information

Potential Federations
Decentralized School
School Systems
State schools, local school districts, etc.
Regional consortia: GigaPoP / *REN
National consortia: Internet2
International: EduRoam
Government: ESNet, NSF, NASA
Industry

(Brief) History of Wireless Security
No RF security
WEP: RC4
easily broken
WPA: TKIP/RC4
many client, AP implementations
WPA2 / 802.11i: CCMP/AES
lacking client implementations
If deploying RF security, WPA as minimum

Focused on 802.1x only?
Concentrate group resources on single strategy
Focus on standards-based solution that would provide a single interface for users
Enables authn, encryption at edge
If necessary, infrastructure could likely be used for non-802.1x

What about Wired?
802.1x on wired is easier than wireless, so it all just works (no active roaming).
We’ve just been saying wireless because it gets attention..

FWNA Project Plan
Work divided in two phases
Phase 1: RADIUS Hierarchy
Initial solution to the problem
Develop knowledge of relevant technology
Understand interoperability issues
Relatively straightforward
Exchange RADIUS keys
Interface to existing authn systems using basic RADIUS mechanism

FWNA Phase 2
Phase 2: RADIUS Federation
Leverage existing federations to enable single-hop RADIUS authentications
Enable attribute release through federations
Requires development
Interface with Shibboleth for authn, inter-site signing
Single-hop server identifications

Beyond authentication…
In many cases today, once authenticated all users obtain same level of service
FWNA is about identity discovery
We must be able to separately provision services from authn and attributes:
Technical setup (IP address, QoS, ACL, etc..)
Access policy
Billing

Other Areas of Investigation
Real Time Diagnostics
Determining cause of authn failure
Requires additional inter-domain data exchange
Access Point Roaming
Will cause re-authentication back to home server (additional delay)
Mitigated by 802.11i pre-authentication

FWNA Project Targets
Phase 1
Toplevel RADIUS server in operation: 1Q05
Phase 2
Early experiments: 3Q05
Operational system: 4Q05

Deliverables
Documents
Architecture: 1Q05
Phase 1 Engineering: 2/05
Phase 1 System Documentation: Ongoing
Phase 2 Plan: 2Q05
Phase 1 System

Join the FWNA Group
Biweekly Conference Calls
 Thursday 11am-12pm: Feb 24, Mar 10
 866-411-0013, 0184827
salsa-fwna @ internet2 list
“subscribe salsa-fwna” to sympa @ internet2