| Strawman for a modified section 6.7 for IEEE 802.1XRev | <– Date –> <– Thread –> |
|
From: Jim Burns (jeb |
|
| Date: Wed, 21 Jan 2004 18:29:59 -0500 (EST) | |
Folks,
Below is a strawman for section 6.7 of the IEEE 802.1XRev document.
It attempts to remedy the discrepencies with section 2.4 of IETF RFC 2284bis.
I have incorporated ideas from original 6.7 and a number of folks, thanks to all.
Please send critiques, comments, concerns and I will try to make modifications to satisfy all.
As Tony J. indicated, we are very late in the 802.1XRev cycle and we will have only one official cycle on this so we should do all our cycling here and get it right the first time.
Thanks,
Jim B.
Jim Burns wrote:
Below is a strawman for section 6.7 of the IEEE 802.1XRev document.
It attempts to remedy the discrepencies with section 2.4 of IETF RFC 2284bis.
I have incorporated ideas from original 6.7 and a number of folks, thanks to all.
Please send critiques, comments, concerns and I will try to make modifications to satisfy all.
As Tony J. indicated, we are very late in the 802.1XRev cycle and we will have only one official cycle on this so we should do all our cycling here and get it right the first time.
Thanks,
Jim B.
Modified IEEE 802.1XREV Section 6.7 Please use a fixed width font to view this. Also, you may need to widen your window to view the table. ---------------------------------------------------- 6.7 Coupling two 802.1X authentications This section discusses the asymmetric nature of the 802.1X transport, how this asymmetry affects authentication results, the ability of the 802.1X transport mechanism to run two simultaneous transports in opposite directions, how this affects authentication results and why this simultaneous bi-directional transport may be utilized.
802.1X's transport mechanism for authentication frames is asymmetric; there is a requestor role (authenticator) and a responder role (supplicant). The authenticator transmits frames to the supplicant and the supplicant responds to those frames with frames of its own. It is a request/response transport and the state machines reflect this. This request/response transport mirrors the authentication protocol that is carried within the 802.1X transport: EAP. Such a transport is sometimes called 'one-way' meaning the protocol is always originating from one side (the requestor).
This transport asymmetry (of 1X and EAP) does not affect the symmetry (or asymmetry) of the authentication method (EAP method) that is carried by the transport. Please note the separate usage of 'authentication protocol (EAP)' from 'authentication method (EAP method)' in this discussion (for a deeper understanding please see RFC 2284). A mutually authenticating EAP method (such as EAP-TLS) may run atop the asymmetric transport mechanism and mutually authenticate both parties (it is symmetric in that both parties contribute to the authentication result). Likewise, if the EAP-method that is chosen is assymmetric (EAP-MD5), then the authentication will be assymmetric (only one party will be authenticated to the other). In fact, it is not the 802.1X transport that controls the symmetry of the authentication, but rather the chosen EAP method and the existence of a controlled port on both the supplicant and authenticator.
It should be noted that in an earlier version of 802.1X (802.1X 2001), only the authenticator was specified to have a controlled port. This previous version of 802.1X did in fact restrict the authentication result to be asymmetric (no matter what EAP method was chosen), but it was due to the lack of a controlled port on the supplicant, not due to the asymmetric transport. The current version of 802.1X remedies this by specifying a controlled port on the supplicant.
The 802.1X transport asymmetry means that each role has its own state machine and it is possible that a single device can implement both roles (state machines). If a device with both roles implemented encounters another device with both roles then two separate (and simultaneous) 802.1X transports will take place in opposite directions. This is sometimes referred to as a bi-directional authentication transport or two coupled one-way authentication transports.
One must understand that two coupled one-way authentication transports is not equivalent in security to a single transport of a mutually authenticating EAP method (such as EAP-TLS). This is because the authentications on the two separate transports are not specified to be cryptographically bound together. In other words there is no way to cryptographically ascertain that the party involved in the first one-way transport is the same party that is involved in the alternate one-way transport.
Despite the ability to mutually authenticate two devices with a single transport, there may exist other reasons to simultaneously run two 802.1X transports in opposite directions. When both devices require the creation of separate keying material as in unicast keys for 802.11 adhoc networks. When different credentials are used for different roles (one credential for the supplicant role and one for the authenticator) then a bi-directional transport with EAP-TLS using different certificates in each direction may be used. When two bridge devices are connected together and they wish to authenticate each other then they may implement both roles on their connecting ports. Some of these cases are discussed in section 2.4 of RFC 2284.
Certain encounters between a device with both .1X transport roles implemented and a device with a single .1X transport role can result in an asymmetric traffic flow (traffic can only flow in one direction). This occurs because one device opens its controlled port while the other device does not. A complete table of encounters and results (assuming authentication successes) appears in Table xxxx.
Table xxxx: Result of encounters between .1X transport roles assuming key-generating EAP methods are run and result in success.
Remote\Local--> | Supplicant | Authenticator | Both | | | | V | | | ---------------------------------------------------------------------- Supplicant |Fails gracefully | Works | Works |No link will be | Authenticated | Authenticated |formed if media | link | link |is considered | Authentication | Authentication |unsafe without | policy is set by | policy set by |encryption | choice of EAP | choice of EAP |(802.11). | method. | method | | | Remote |Unauthenticated | | supplicant and |link will be | | local |formed if media | | authenticator |is considered | | are satisfied. |safe by default | | Local |(802.3) | | supplicant will | | | receive no | | | response and | | | assume no | | | authenticator. | | | Successful | | | authentication | | | to the local | | | authenticator | | | will have made | | | the port | | | secure. ----------------------------------------------------------------------- Authenticator |Works - |Fails gracefully |Fails – |Authenticated |No link will be |Asymmetric link |link – |formed. Each |Remote |Authentication |authenticator |controlled port |policy set by |will send initial |is operational, |choice of EAP |EAP requests but |but local |method. |no response will |controlled port | |arrive. |remains non- | | |operational. | | |Remote | | |authenticator | | |and local | | |supplicant are | | |satisfied, | | |local | | |authenticator | | |remains | | |unsatisfied. ----------------------------------------------------------------------- Both |Works |Fails |Works |Authenticated |Asymmetric link | Authenticated |link |Local controlled | link |Authentication |port is | Authentication |policy set by |operational, but | policy set by |choice of EAP |remote controlled | choice of two |method. |port remains non- | uncoupled EAP |Local supplicant |operational. | methods. Two |and remote |Local | keys will be |authenticator |authenticator and | formed and some |are satisfied. |remote supplicant | mechanism must |Remote |are satisfied, | exist so that |supplicant will |remote | local and |receive no |authenticator | remote choose |response and |remains | the same key. |assume no |unsatisfied. | |authenticator. | | |Successful | | |authentication | | |to the remote | | |authenticator | | |will have made | | |the port secure. | | -----------------------------------------------------------------------
Applications that utilize 802.1X should clearly specify which .1X roles must be implemented to avoid encounters that will result in failed network connections. In addition, an application should consider specifying a class of EAP methods or a limited set of EAP methods to allow for simpler security analysis of the application. Some examples of known applications and their common usage of 802.1X and EAP are:
802.11i infrastructure mode - Access point implements authenticator role - Client implements the supplicant role - A mutually authenticating EAP method is utilized
802.1 bridge to end node - Bridge implements the authenticator role - End node implements the supplicant role - No EAP method type specified
802.11 adhoc mode - Each adhoc device implements both the supplicant and authenticator role - A key-generating EAP method is utilized.
802.1 bridge to 802.1 bridge - Each bridge implements both the supplicant and authenticator role. - Each bridge sets the Supplicant Access Control With Authenticator administrative control parameter to inactive so that only the authenticator role controls the controlled port.
Jim Burns wrote:
Hi Bernard,
Yes, I would be happy to take a first crack at it and then accept input.
My schedule is such that I will not get it out til this eve or Wednesday AM.
Sorry about the delay,
Jim B.
Bernard Aboba wrote:
Hi folks, Yes, a reread of 6.7 leads me to believe that it could be confusing to implementers. My opinion is that the current 6.7 was written before there was a supplicant controlled port. With the addition of the supplicant controlled port, the uni-directional nature of the transport is no longer a blocking feature to a bi-directional EAP method.
I agree (modulo the caveats in Section 2.4 of RFC 2284bis).
I believe the points we want to make in 6.7 are the following:
This sounds like a good start. Do you want to draft some sample text for review?
=>IEEE 802.1 Email List user information: http://www.ieee802.org/1/email-pages/
- Re: Discrepancies between 802.1XREV and RFC 2248bis, (continued)
- Re: Discrepancies between 802.1XREV and RFC 2248bis Bernard Aboba, January 19 2004
-
Re: Discrepancies between 802.1XREV and RFC 2248bis Jim Burns, January 19 2004
-
Re: Discrepancies between 802.1XREV and RFC 2248bis Bernard Aboba, January 19 2004
- Re: Discrepancies between 802.1XREV and RFC 2248bis Jim Burns, January 20 2004
- Strawman for a modified section 6.7 for IEEE 802.1XRev Jim Burns, January 21 2004
-
Re: Discrepancies between 802.1XREV and RFC 2248bis Bernard Aboba, January 19 2004
- Re: Discrepancies between 802.1XREV and RFC 2248bis Jari Arkko, January 21 2004
Results generated by Tiger Technologies using MHonArc.