RFC 7831






Internet Engineering Task Force (IETF)                        J. Howlett
Request for Comments: 7831                                          Jisc
Category: Informational                                       S. Hartman
ISSN: 2070-1721                                        Painless Security
                                                           H. Tschofenig
                                                                ARM Ltd.
                                                               J. Schaad
                                                          August Cellars
                                                                May 2016


      Application Bridging for Federated Access Beyond Web (ABFAB)
                              Architecture

Abstract



   Over the last decade, a substantial amount of work has occurred in
   the space of federated access management.  Most of this effort has
   focused on two use cases: network access and web-based access.
   However, the solutions to these use cases that have been proposed and
   deployed tend to have few building blocks in common.

   This memo describes an architecture that makes use of extensions to
   the commonly used security mechanisms for both federated and non-
   federated access management, including the Remote Authentication
   Dial-In User Service (RADIUS), the Generic Security Service
   Application Program Interface (GSS-API), the Extensible
   Authentication Protocol (EAP), and the Security Assertion Markup
   Language (SAML).  The architecture addresses the problem of federated
   access management to primarily non-web-based services, in a manner
   that will scale to large numbers of Identity Providers, Relying
   Parties, and federations.

Status of This Memo



   This document is not an Internet Standards Track specification; it is
   published for informational purposes.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Not all documents
   approved by the IESG are a candidate for any level of Internet
   Standard; see Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc7831.



Howlett, et al.               Informational                     [Page 1]

RFC 7831                   ABFAB Architecture                   May 2016


Copyright Notice



   Copyright (c) 2016 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.





































Howlett, et al.               Informational                     [Page 2]

RFC 7831                   ABFAB Architecture                   May 2016


Table of Contents



   1. Introduction ....................................................4
      1.1. Terminology ................................................5
           1.1.1. Channel Binding .....................................6
      1.2. An Overview of Federation ..................................8
      1.3. Challenges for Contemporary Federation ....................11
      1.4. An Overview of ABFAB-Based Federation .....................11
      1.5. Design Goals ..............................................14
   2. Architecture ...................................................15
      2.1. Relying Party to Identity Provider ........................16
           2.1.1. AAA, RADIUS, and Diameter ..........................17
           2.1.2. Discovery and Rules Determination ..................19
           2.1.3. Routing and Technical Trust ........................20
           2.1.4. AAA Security .......................................21
           2.1.5. SAML Assertions ....................................22
      2.2. Client to Identity Provider ...............................24
           2.2.1. Extensible Authentication Protocol (EAP) ...........24
           2.2.2. EAP Channel Binding ................................26
      2.3. Client to Relying Party ...................................26
           2.3.1. GSS-API ............................................27
           2.3.2. Protocol Transport .................................28
           2.3.3. Re-authentication ..................................29
   3. Application Security Services ..................................29
      3.1. Authentication ............................................29
      3.2. GSS-API Channel Binding ...................................31
      3.3. Host-Based Service Names ..................................32
      3.4. Additional GSS-API Services ...............................33
   4. Privacy Considerations .........................................34
      4.1. Entities and Their Roles ..................................35
      4.2. Privacy Aspects of ABFAB Communication Flows ..............36
           4.2.1. Client to RP .......................................36
           4.2.2. Client to IdP (via Federation Substrate) ...........37
           4.2.3. IdP to RP (via Federation Substrate) ...............38
      4.3. Relationship between User and Entities ....................39
      4.4. Accounting Information ....................................39
      4.5. Collection and Retention of Data and Identifiers ..........39
      4.6. User Participation ........................................40
   5. Security Considerations ........................................40
   6. References .....................................................41
      6.1. Normative References ......................................41
      6.2. Informative References ....................................42
   Acknowledgments ...................................................46
   Authors' Addresses ................................................46







Howlett, et al.               Informational                     [Page 3]

RFC 7831                   ABFAB Architecture                   May 2016


1.  Introduction



   Numerous security mechanisms have been deployed on the Internet to
   manage access to various resources.  These mechanisms have been
   generalized and scaled over the last decade through mechanisms such
   as the Simple Authentication and Security Layer (SASL) with the
   Generic Security Server Application Program Interface (GSS-API)
   (known as the GS2 family) [RFC5801]; the Security Assertion Markup
   Language (SAML) [OASIS.saml-core-2.0-os]; and the Authentication,
   Authorization, and Accounting (AAA) architecture as embodied in
   RADIUS [RFC2865] and Diameter [RFC6733].

   A Relying Party (RP) is the entity that manages access to some
   resource.  The entity that is requesting access to that resource is
   often described as the client.  Many security mechanisms are
   manifested as an exchange of information between these entities.
   The RP is therefore able to decide whether the client is authorized
   or not.

   Some security mechanisms allow the RP to delegate aspects of the
   access management decision to an entity called the Identity Provider
   (IdP).  This delegation requires technical signaling, trust, and a
   common understanding of semantics between the RP and IdP.  These
   aspects are generally managed within a relationship known as a
   "federation".  This style of access management is accordingly
   described as "federated access management".

   Federated access management has evolved over the last decade through
   specifications like SAML [OASIS.saml-core-2.0-os], OpenID
   (http://www.openid.net), OAuth [RFC6749], and WS-Trust [WS-TRUST].
   The benefits of federated access management include:

   Single or simplified sign-on:

      An Internet service can delegate access management, and the
      associated responsibilities such as identity management and
      credentialing, to an organization that already has a long-term
      relationship with the client.  This is often attractive, as RPs
      frequently do not want these responsibilities.  The client also
      requires fewer credentials, which is also desirable.











Howlett, et al.               Informational                     [Page 4]

RFC 7831                   ABFAB Architecture                   May 2016


   Data minimization and user participation:

      Often, an RP does not need to know the identity of a client to
      reach an access management decision.  It is frequently only
      necessary for the RP to know specific attributes about the client
      -- for example, that the client is affiliated with a particular
      organization or has a certain role or entitlement.  Sometimes, the
      RP only needs to know a pseudonym of the client.

      Prior to the release of attributes to the RP from the IdP, the IdP
      will check configuration and policy to determine if the attributes
      are to be released.  There is currently no direct client
      participation in this decision.

   Provisioning:

      Sometimes, an RP needs, or would like, to know more about a client
      than an affiliation or a pseudonym.  For example, an RP may want
      the client's email address or name.  Some federated access
      management technologies provide the ability for the IdP to supply
      this information, either on request by the RP or unsolicited.

   This memo describes the Application Bridging for Federated Access
   Beyond web (ABFAB) architecture.  This architecture addresses the
   problem of federated access management primarily for non-web-based
   services.  This architecture makes use of extensions to the commonly
   used security mechanisms for both federated and non-federated access
   management, including RADIUS, the Generic Security Service (GSS), the
   Extensible Authentication Protocol (EAP), and SAML.  The architecture
   should be extended to use Diameter in the future.  It does so in a
   manner that is designed to scale to large numbers of IdPs, RPs, and
   federations.

1.1.  Terminology



   This document uses identity management and privacy terminology from
   [RFC6973].  In particular, this document uses the terms
   "identity provider", "relying party", "identifier", "pseudonymity",
   "unlinkability", and "anonymity".

   In this architecture, the IdP consists of the following components:
   an EAP server, a RADIUS server, and, optionally, a SAML Assertion
   service.

   This document uses the term "Network Access Identifier" (NAI) as
   defined in [RFC7542].  An NAI consists of a realm identifier, which
   is associated with a AAA server, and thus an IdP and a username, that
   are associated with a specific client of the IdP.



Howlett, et al.               Informational                     [Page 5]

RFC 7831                   ABFAB Architecture                   May 2016


   One of the problems some people have found with reading this document
   is that the terminology sometimes appears to be inconsistent.  This
   is because the various standards that we refer to use different terms
   for the same concept.  In general, this document uses either the
   ABFAB term or the term associated with the standard under discussion,
   as appropriate.  For reference, we include Table 1 below, which
   provides a mapping for these different terms.  (Note that items
   marked "N/A" (not applicable) indicate that there is no name that
   represents the entity.)

   +----------+-----------+--------------------+-----------------------+
   | Protocol | Client    | Relying Party      | Identity Provider     |
   +----------+-----------+--------------------+-----------------------+
   | ABFAB    | N/A       | Relying Party (RP) | Identity Provider     |
   |          |           |                    | (IdP)                 |
   |          |           |                    |                       |
   |          | Initiator | Acceptor           | N/A                   |
   |          |           |                    |                       |
   |          | Client    | Server             | N/A                   |
   |          |           |                    |                       |
   | SAML     | Subject   | Service provider   | Issuer                |
   |          |           |                    |                       |
   | GSS-API  | Initiator | Acceptor           | N/A                   |
   |          |           |                    |                       |
   | EAP      | EAP peer  | EAP authenticator  | EAP server            |
   |          |           |                    |                       |
   | AAA      | N/A       | AAA client         | AAA server            |
   |          |           |                    |                       |
   | RADIUS   | user      | NAS                | N/A                   |
   |          |           |                    |                       |
   |          | N/A       | RADIUS client      | RADIUS server         |
   +----------+-----------+--------------------+-----------------------+

                           Table 1: Terminology

1.1.1.  Channel Binding



   This document uses the term "channel binding" in two different
   contexts; this term has a different meaning in each of these
   contexts.

   EAP channel binding is used to implement GSS-API naming semantics.
   EAP channel binding sends a set of attributes from the peer to the
   EAP server either as part of the EAP conversation or as part of a
   secure association protocol.  In addition, attributes are sent in the
   back-end protocol from the EAP authenticator to the EAP server.  The





Howlett, et al.               Informational                     [Page 6]

RFC 7831                   ABFAB Architecture                   May 2016


   EAP server confirms the consistency of these attributes and provides
   the confirmation back to the peer.  In this document, channel binding
   without qualification refers to EAP channel binding.

   GSS-API channel binding provides protection against man-in-the-middle
   attacks when GSS-API is used for authentication inside of some
   tunnel; it is similar to a facility called "cryptographic binding" in
   EAP.  The binding works by each side deriving a cryptographic value
   from the tunnel itself and then using that cryptographic value to
   prove to the other side that it knows the value.

   See [RFC5056] for a discussion of the differences between these two
   facilities.  These differences can be summarized as follows:

   o  GSS-API channel binding specifies that there is nobody between the
      client and the EAP authenticator.

   o  EAP channel binding allows the client to have knowledge of such
      EAP authenticator attributes as the EAP authenticator's name.

   Typically, when considering both EAP and GSS-API channel binding,
   people think of channel binding in combination with mutual
   authentication.  This is sufficiently common that, without additional
   qualification, channel binding should be assumed to imply mutual
   authentication.  In GSS-API, without mutual authentication, only the
   acceptor has authenticated the initiator.  Similarly, in EAP, only
   the EAP server has authenticated the peer.  Sometimes, one-way
   authentication is useful.  Consider, for example, a user who wishes
   to access a protected resource for a shared whiteboard in a
   conference room.  The whiteboard is the acceptor; it knows that the
   initiator is authorized to give it a presentation, and the user can
   validate that the whiteboard got the correct presentation by visual
   means.  (The presentation should not be confidential in this case.)
   If channel binding is used without mutual authentication, it is
   effectively a request to disclose the resource in the context of a
   particular channel.  Such an authentication would be similar in
   concept to a holder-of-key SAML Assertion.  However, note also that
   although it is not happening in the protocol, mutual authentication
   is happening in the overall system: the user is able to visually
   authenticate the content.  This is consistent with all uses of
   channel binding without protocol-level mutual authentication found
   so far.









Howlett, et al.               Informational                     [Page 7]

RFC 7831                   ABFAB Architecture                   May 2016


1.2.  An Overview of Federation



   In the previous section, we introduced the following entities:

   o  the client,

   o  the IdP, and

   o  the RP.

   The final entity that needs to be introduced is the Individual.  An
   Individual is a human being that is using the client.  In any given
   situation, an Individual may or may not exist.  Clients can act as
   front ends for Individuals, or clients may be independent entities
   that are set up and allowed to run autonomously.  An example of such
   an independent entity can be found in the Trust Router Protocol
   (https://www.ietf.org/proceedings/86/slides/slides-86-rtgarea-0.pdf),
   where the routers use ABFAB to authenticate to each other.

   These entities and their relationships are illustrated graphically in
   Figure 1.

             ,----------\                        ,---------\
             | Identity |       Federation       | Relying |
             | Provider + <--------------------> + Party   |
             `----------'                        '---------'
                    <
                     \
                      \ Authentication
                       \
                        \
                         \
                          \
                           \  +---------+
                            \ |         |  O
                             v| Client  | \|/ Individual
                              |         |  |
                              +---------+ / \

                Figure 1: Entities and Their Relationships











Howlett, et al.               Informational                     [Page 8]

RFC 7831                   ABFAB Architecture                   May 2016


   The relationships between the entities in Figure 1 are as follows:

   Federation

      The IdP and the RPs are part of a federation.  The relationship
      may be direct (they have an explicit trust relationship) or
      transitive (the trust relationship is mediated by one or more
      entities).  The federation relationship is governed by a
      federation agreement.  Within a single federation, there may be
      multiple IdPs as well as multiple RPs.

   Authentication

      There is a direct relationship between the client and the IdP.
      This relationship provides the means by which they trust each
      other and can securely authenticate each other.

   A federation agreement typically encompasses operational
   specifications and legal rules:

   Operational Specifications:

      The goal of operational specifications is to provide enough
      definition that the system works and interoperability is possible.
      These include the technical specifications (e.g., protocols used
      to communicate between the three parties), process standards,
      policies, identity proofing, credential and authentication
      algorithm requirements, performance requirements, assessment and
      audit criteria, etc.

   Legal Rules:

      The legal rules take the legal framework into consideration and
      provide contractual obligations for each entity.  The rules define
      the responsibilities of each party and provide further
      clarification of the operational specifications.  These legal
      rules regulate the operational specifications, make operational
      specifications legally binding to the participants, and define and
      govern the rights and responsibilities of the participants.  The
      legal rules may, for example, describe liability for losses,
      termination rights, enforcement mechanisms, measures of damage,
      dispute resolution, warranties, etc.









Howlett, et al.               Informational                     [Page 9]

RFC 7831                   ABFAB Architecture                   May 2016


   The operational specifications can demand the usage of a specific
   technical infrastructure, including requirements on the message
   routing intermediaries, to offer the required technical
   functionality.  In other environments, the operational specifications
   require fewer technical components in order to meet the required
   technical functionality.

   The legal rules include many non-technical aspects of federation,
   such as business practices and legal arrangements, which are outside
   the scope of the IETF.  The legal rules can still have an impact on
   the architectural setup or on how to ensure the dynamic establishment
   of trust.

   While a federation agreement is often discussed within the context of
   formal relationships, such as between an enterprise and an employee
   or between a government and a citizen, a federation agreement does
   not have to require any particular level of formality.  For an IdP
   and a client, it is sufficient for a relationship to be established
   by something as simple as using a web form and confirmation email.
   For an IdP and an RP, it is sufficient for the IdP to publish contact
   information along with a public key and for the RP to use that data.
   Within the framework of ABFAB, it will generally be required that a
   mechanism exist for the IdP to be able to trust the identity of the
   RP; if this is not present, then the IdP cannot provide the
   assurances to the client that the identity of the RP has been
   established.

   The nature of federation dictates that there exists some form of
   relationship between the IdP and the RP.  This is particularly
   important when the RP wants to use information obtained from the IdP
   for access management decisions and when the IdP does not want to
   release information to every RP (or only under certain conditions).

   While it is possible to have a bilateral agreement between every IdP
   and every RP, on an Internet scale, this setup requires the
   introduction of the multilateral federation concept, as the
   management of such pair-wise relationships would otherwise prove
   burdensome.

   The IdP will typically have a long-term relationship with the client.
   This relationship typically involves the IdP positively identifying
   and credentialing the client (for example, at the time of employment
   within an organization).  When dealing with Individuals, this process
   is called "identity proofing" [NIST-SP.800-63-2].  The relationship
   will often be instantiated within an agreement between the IdP and
   the client (for example, within an employment contract or terms of
   use that stipulate the appropriate use of credentials and so forth).




Howlett, et al.               Informational                    [Page 10]

RFC 7831                   ABFAB Architecture                   May 2016


   The nature and quality of the relationship between the client and the
   IdP are important contributors to the level of trust that an RP may
   assign to an assertion describing a client made by an IdP.  This is
   sometimes described as the level of assurance [NIST-SP.800-63-2].

   Federation does not require an a priori relationship or a long-term
   relationship between the RP and the client; it is this property of
   federation that yields many of its benefits.  However, federation
   does not preclude the possibility of a pre-existing relationship
   between the RP and the client or the possibility that the RP and
   client may use the introduction to create a new long-term
   relationship independent of the federation.

   Finally, it is important to reiterate that in some scenarios there
   might indeed be an Individual behind the client and in other cases
   the client may be autonomous.

1.3.  Challenges for Contemporary Federation



   As federated IdPs and RPs (services) proliferate, the role of an
   Individual can become ambiguous in certain circumstances.  For
   example, a school might provide online access for a student's grades
   to their parents for review and to the student's teacher for
   modification.  A teacher who is also a parent must clearly
   distinguish their role upon access.

   Similarly, as federations proliferate, it becomes increasingly
   difficult to discover which IdP(s) a user is associated with.  This
   is true for both the web and non-web case but is particularly acute
   for the latter, as many non-web authentication systems are not
   semantically rich enough on their own to allow for such ambiguities.
   For instance, in the case of an email provider, SMTP and IMAP do not
   have the ability for the server to request information from the
   client, beyond the client NAI, that the server would then use to
   decide between the multiple federations it is associated with.
   However, the building blocks do exist to add this functionality.

1.4.  An Overview of ABFAB-Based Federation



   The previous section described the general model of federation and
   the application of access management within the federation.  This
   section provides a brief overview of ABFAB in the context of this
   model.








Howlett, et al.               Informational                    [Page 11]

RFC 7831                   ABFAB Architecture                   May 2016


   In this example, a client is attempting to connect to a server in
   order to either get access to some data or perform some type of
   transaction.  In order for the client to mutually authenticate with
   the server, the following steps are taken in an ABFAB architecture (a
   graphical view of the steps can be found in Figure 2):

   1.   Client configuration: The client is configured with an NAI
        assigned by the IdP.  It is also configured with any keys,
        certificates, passwords, or other secret and public information
        needed to run the EAP protocols between it and the IdP.

   2.   Authentication mechanism selection: The client is configured to
        use the GSS-EAP GSS-API mechanism for authentication/
        authorization.



   3.   Client provides an NAI to RP: The client sets up a transport to
        the RP and begins GSS-EAP authentication.  In response, the RP
        sends an EAP request message (nested in GSS-EAP) asking for the
        client's name.  The client sends an EAP response with an NAI
        name form that, at a minimum, contains the realm portion of its
        full NAI.

   4.   Discovery of federated IdP: The RP uses preconfigured
        information or a federation proxy to determine what IdP to use,
        based on policy and the realm portion of the provided client
        NAI.  This is discussed in detail below (Section 2.1.2).

   5.   Request from RP to IdP: Once the RP knows who the IdP is, it (or
        its agent) will send a RADIUS request to the IdP.  The RADIUS
        Access-Request encapsulates the EAP response.  At this stage,
        the RP will likely have no idea who the client is.  The RP sends
        its identity to the IdP in AAA attributes, and it may send a
        SAML request in a AAA attribute.  The AAA network checks to see
        that the identity claimed by the RP is valid.

   6.   IdP begins EAP with the client: The IdP sends an EAP message to
        the client with an EAP method to be used.  The IdP should not
        re-request the client's name in this message, but clients need
        to be able to handle it.  In this case, the IdP must accept a
        realm only in order to protect the client's name from the RP.
        The available and appropriate methods are discussed below
        (Section 2.2.1).

   7.   EAP is run: A bunch of EAP messages are passed between the
        client (EAP peer) and the IdP (EAP server), until the result of
        the authentication protocol is determined.  The number and
        content of those messages depend on the EAP method selected.  If
        the IdP is unable to authenticate the client, the IdP sends an



Howlett, et al.               Informational                    [Page 12]

RFC 7831                   ABFAB Architecture                   May 2016


        EAP failure message to the RP.  As part of the EAP method, the
        client sends an EAP channel-binding message to the IdP
        (Section 2.2.2).  In the channel-binding message, the client
        identifies, among other things, the RP to which it is attempting
        to authenticate.  The IdP checks the channel-binding data from
        the client against the data provided by the RP via the AAA
        protocol.  If the bindings do not match, the IdP sends an EAP
        failure message to the RP.

   8.   Successful EAP authentication: At this point, the IdP (EAP
        server) and client (EAP peer) have mutually authenticated each
        other.  As a result, the client and the IdP hold two
        cryptographic keys: a Master Session Key (MSK) and an Extended
        MSK (EMSK).  At this point, the client has a level of assurance
        regarding the identity of the RP, based on the name checking the
        IdP has done, using the RP naming information from the AAA
        framework and from the client (by the channel-binding data).

   9.   Local IdP policy check: At this stage, the IdP checks local
        policy to determine whether the RP and client are authorized for
        a given transaction/service and, if so, what attributes, if any,
        will be released to the RP.  If the IdP gets a policy failure,
        it sends an EAP failure message to the RP and client.  (The RP
        will have done its policy checks during the discovery process.)

   10.  IdP provides the RP with the MSK: The IdP sends a success result
        EAP to the RP, along with an optional set of AAA attributes
        associated with the client (usually as one or more SAML
        Assertions).  In addition, the EAP MSK is returned to the RP.

   11.  RP processes results: When the RP receives the result from the
        IdP, it should have enough information to either grant or refuse
        a resource Access-Request.  It may have information that
        associates the client with specific authorization identities.
        If additional attributes are needed from the IdP, the RP may
        make a new SAML request to the IdP.  It will apply these results
        in an application-specific way.

   12.  RP returns results to client: Once the RP has a response, it
        must inform the client of the result.  If all has gone well, all
        are authenticated, and the application proceeds with appropriate
        authorization levels.  The client can now complete the
        authentication of the RP by using the EAP MSK value.








Howlett, et al.               Informational                    [Page 13]

RFC 7831                   ABFAB Architecture                   May 2016


        Relying         Client         Identity
        Party                          Provider

        |              (1)             | Client configuration
        |               |              |
        |<-----(2)----->|              | Mechanism selection
        |               |              |
        |<-----(3)-----<|              | NAI transmitted to RP
        |               |              |
        |<=====(4)====================>| IdP Discovery
        |               |              |
        |>=====(5)====================>| Access-Request from RP to IdP
        |               |              |
        |               |< - - (6) - -<| EAP method to client
        |               |              |
        |               |< - - (7) - ->| EAP exchange to authenticate
        |               |              | client
        |               |              |
        |               |           (8 & 9) Local policy check
        |               |              |
        |<====(10)====================<| Results to RP
        |               |              |
      (11)              |              | RP processes results
        |               |              |
        |>----(12)----->|              | Results to client

        Legend:

          -----: Between client and RP
          =====: Between RP and IdP
          - - -: Between client and IdP (via RP)

                   Figure 2: ABFAB Authentication Steps

1.5.  Design Goals



   Our key design goals are as follows:

   o  Each party in a transaction will be authenticated, although
      perhaps not identified, and the client will be authorized for
      access to a specific resource.

   o  The means of authentication is decoupled from the application
      protocol so as to allow for multiple authentication methods with
      minimal changes to the application.

   o  The architecture requires no sharing of long-term private keys
      between clients and RPs.



Howlett, et al.               Informational                    [Page 14]

RFC 7831                   ABFAB Architecture                   May 2016


   o  The system will scale to large numbers of IdPs, RPs, and users.

   o  The system will be designed primarily for non-web-based
      authentication.

   o  The system will build upon existing standards, components, and
      operational practices.

   Designing new three-party authentication and authorization protocols
   is difficult and fraught with the risk of cryptographic flaws.
   Achieving widespread deployment is even more difficult.  A lot of
   attention on federated access has been devoted to the web.  This
   document instead focuses on a non-web-based environment and focuses
   on those protocols where HTTP is not used.  Despite the growing trend
   to layer every protocol on top of HTTP, there are still a number of
   protocols available that do not use HTTP-based transports.  Many of
   these protocols are lacking a native authentication and authorization
   framework of the style shown in Figure 1.

2.  Architecture

   We have already introduced the federated access architecture, with
   the illustration of the different actors that need to interact.  This
   section expands on the specifics of providing support for
   non-web-based applications and provides motivations for design
   decisions.  The main theme of the work described in this document is
   focused on reusing existing building blocks that have been deployed
   already and to rearrange them in a novel way.

   Although this architecture assumes updates to the RP, the client, and
   the IdP, those changes are kept at a minimum.  A mechanism that can
   demonstrate deployment benefits (based on ease of updates to existing
   software, low implementation effort, etc.) is preferred, and there
   may be a need to specify multiple mechanisms to support the range of
   different deployment scenarios.

   There are a number of ways to encapsulate EAP into an application
   protocol.  For ease of integration with a wide range of non-web-based
   application protocols, GSS-API was chosen.  The technical
   specification of GSS-EAP can be found in [RFC7055].

   The architecture consists of several building blocks, as shown
   graphically in Figure 3.  In the following sections, we discuss the
   data flow between each of the entities, the protocols used for that
   data flow, and some of the trade-offs made in choosing the protocols.






Howlett, et al.               Informational                    [Page 15]

RFC 7831                   ABFAB Architecture                   May 2016


                                    +--------------+
                                    |   Identity   |
                                    |   Provider   |
                                    |    (IdP)     |
                                    +-^----------^-+
                                      * EAP      o RADIUS
                                      *          o
                                    --v----------v--
                                 ///                \\\
                               //                      \\
                              |        Federation        |
                              |        Substrate         |
                               \\                      //
                                 \\\                ///
                                    --^----------^--
                                      * EAP      o RADIUS
                                      *          o
   +-------------+                  +-v----------v--+
   |             |                  |               |
   | Client      |  EAP/EAP Method  | Relying Party |
   | Application |<****************>|     (RP)      |
   |             |  GSS-API         |               |
   |             |<---------------->|               |
   |             |  Application     |               |
   |             |  Protocol        |               |
   |             |<================>|               |
   +-------------+                  +---------------+

   Legend:

     <****>: Client-to-IdP Exchange
     <---->: Client-to-RP Exchange
     <oooo>: RP-to-IdP Exchange
     <====>: Protocol through which GSS-API/GS2 exchanges are tunneled

                  Figure 3: ABFAB Protocol Instantiation

2.1.  Relying Party to Identity Provider



   Communication between the RP and the IdP is done by the Federation
   Substrate.  This communication channel is responsible for:

   o  Establishing the trust relationship between the RP and the IdP.

   o  Determining the rules governing the relationship.

   o  Conveying authentication packets from the client to the IdP
      and back.



Howlett, et al.               Informational                    [Page 16]

RFC 7831                   ABFAB Architecture                   May 2016


   o  Providing the means of establishing a trust relationship between
      the RP and the client.

   o  Providing a means for the RP to obtain attributes about the client
      from the IdP.

   The ABFAB working group has chosen the AAA framework for the messages
   transported between the RP and IdP.  The AAA framework supports the
   requirements stated above, as follows:

   o  The AAA backbone supplies the trust relationship between the RP
      and the IdP.

   o  The agreements governing a specific AAA backbone contain the rules
      governing the relationships within the AAA federation.

   o  A method exists for carrying EAP packets within RADIUS [RFC3579]
      and Diameter [RFC4072].

   o  The use of EAP channel binding [RFC6677] along with the core ABFAB
      protocol provide the pieces necessary to establish the identities
      of the RP and the client, while EAP provides the cryptographic
      methods for the RP and the client to validate that they are
      talking to each other.

   o  A method exists for carrying SAML packets within RADIUS [RFC7833];
      this method allows the RP to query attributes about the client
      from the IdP.

   Protocols that support the same framework but do different routing
   are expected to be defined and used in the future.  One such effort,
   called the Trust Router, is to set up a framework that creates a
   trusted point-to-point channel on the fly
   (https://www.ietf.org/proceedings/86/slides/slides-86-rtgarea-0.pdf).

2.1.1.  AAA, RADIUS, and Diameter



   The usage of the AAA framework with RADIUS [RFC2865] and Diameter
   [RFC6733] for network access authentication has been successful from
   a deployment point of view.  To map the terminology used in Figure 1
   to the AAA framework, the IdP corresponds to the AAA server; the RP
   corresponds to the AAA client; and the technical building blocks of a
   federation are AAA proxies, relays, and redirect agents (particularly
   if they are operated by third parties, such as AAA brokers and
   clearinghouses).  In the case of network access authentication, the
   front end, i.e., the communication path between the end host and the
   AAA client, is offered by link-layer protocols that forward




Howlett, et al.               Informational                    [Page 17]

RFC 7831                   ABFAB Architecture                   May 2016


   authentication protocol exchanges back and forth.  An example of a
   large-scale RADIUS-based federation is eduroam
   (https://www.eduroam.org).

   By using the AAA framework, ABFAB can be built on the federation
   agreements that already exist; the agreements can then merely be
   expanded to cover the ABFAB architecture.  The AAA framework has
   already addressed some of the problems outlined above.  For example,

   o  It already has a method for routing requests based on a domain.

   o  It already has an extensible architecture allowing for new
      attributes to be defined and transported.

   o  Pre-existing relationships can be reused.

   The astute reader will notice that RADIUS and Diameter have
   substantially similar characteristics.  Why not pick one?  RADIUS and
   Diameter are deployed in different environments.  RADIUS can often be
   found in enterprise and university networks; RADIUS is also used by
   operators of fixed networks.  Diameter, on the other hand, is
   deployed by operators of mobile networks.  Another key difference is
   that today RADIUS is largely transported over UDP.  The decision
   regarding which protocol will be appropriate to deploy is left to
   implementers.  The protocol defines all the necessary new AAA
   attributes as RADIUS attributes.  A future document could define the
   same AAA attributes for a Diameter environment.  We also note that
   there exist proxies that convert from RADIUS to Diameter and back.
   This makes it possible for both to be deployed in a single Federation
   Substrate.

   Through the integrity-protection mechanisms in the AAA framework, the
   IdP can establish technical trust that messages are being sent by the
   appropriate RP.  Any given interaction will be associated with one
   federation at the policy level.  The legal or business relationship
   defines what statements the IdP is trusted to make and how these
   statements are interpreted by the RP.  The AAA framework also permits
   the RP or elements between the RP and IdP to make statements about
   the RP.

   The AAA framework provides transport for attributes.  Statements made
   about the client by the IdP, statements made about the RP, and other
   information are transported as attributes.

   One demand that the AAA substrate makes of the upper layers is that
   they must properly identify the endpoints of the communication.  It
   must be possible for the AAA client at the RP to determine where to
   send each RADIUS or Diameter message.  Without this requirement, it



Howlett, et al.               Informational                    [Page 18]

RFC 7831                   ABFAB Architecture                   May 2016


   would be the RP's responsibility to determine the identity of the
   client on its own, without the assistance of an IdP.  This
   architecture makes use of the Network Access Identifier (NAI), where
   the IdP is indicated by the realm component [RFC7542].  The NAI is
   represented and consumed by the GSS-API layer as GSS_C_NT_USER_NAME,
   as specified in [RFC2743].  The GSS-API EAP mechanism includes the
   NAI in the EAP Response/Identity message.

   At the time of this writing, no profiles for the use of Diameter have
   been created.

2.1.2.  Discovery and Rules Determination



   While we are using the AAA protocols to communicate with the IdP, the
   RP may have multiple Federation Substrates to select from.  The RP
   has a number of criteria that it will use in selecting which of the
   different federations to use.  The federation selected must

   o  be able to communicate with the IdP.

   o  match the business rules and technical policies required for the
      RP security requirements.

   The RP needs to discover which federation will be used to contact the
   IdP.  The first selection criterion used during discovery is going to
   be the name of the IdP to be contacted.  The second selection
   criterion used during discovery is going to be the set of business
   rules and technical policies governing the relationship; this is
   called "rules determination".  The RP also needs to establish
   technical trust in the communications with the IdP.

   Rules determination covers a broad range of decisions about the
   exchange.  One of these is whether the given RP is permitted to talk
   to the IdP using a given federation at all, so rules determination
   encompasses the basic authorization decision.  Other factors are
   included, such as what policies govern release of information about
   the client to the RP and what policies govern the RP's use of this
   information.  While rules determination is ultimately a business
   function, it has a significant impact on the technical exchanges.
   The protocols need to communicate the result of authorization.  When
   multiple sets of rules are possible, the protocol must disambiguate
   which set of rules are in play.  Some rules have technical
   enforcement mechanisms; for example, in some federations,
   intermediaries validate information that is being communicated within
   the federation.






Howlett, et al.               Informational                    [Page 19]

RFC 7831                   ABFAB Architecture                   May 2016


   At the time of this writing, no protocol mechanism has been specified
   to allow a AAA client to determine whether a AAA proxy will indeed be
   able to route AAA requests to a specific IdP.  The AAA routing is
   impacted by business rules and technical policies that may be quite
   complex; at the present time, the route selection is based on manual
   configuration.

2.1.3.  Routing and Technical Trust



   Several approaches to having messages routed through the Federation
   Substrate are possible.  These routing methods can most easily be
   classified based on the mechanism for technical trust that is used.
   The choice of technical trust mechanism constrains how rules
   determination is implemented.  Regardless of what deployment strategy
   is chosen, it is important that the technical trust mechanism be able
   to validate the identities of both parties to the exchange.  The
   trust mechanism must ensure that the entity acting as the IdP for a
   given NAI is permitted to be the IdP for that realm and that any
   service name claimed by the RP is permitted to be claimed by that
   entity.  Here are the categories of technical trust determination:

   AAA Proxy:
      The simplest model is that an RP is a AAA client and can send the
      request directly to a AAA proxy.  The hop-by-hop integrity
      protection of the AAA fabric provides technical trust.  An RP can
      submit a request directly to the correct federation.
      Alternatively, a federation disambiguation fabric can be used.
      Such a fabric takes information about what federations the RP is
      part of and what federations the IdP is part of, and it routes a
      message to the appropriate federation.  The routing of messages
      across the fabric, plus attributes added to requests and
      responses, together provide rules determination.  For example,
      when a disambiguation fabric routes a message to a given
      federation, that federation's rules are chosen.  Name validation
      is enforced as messages travel across the fabric.  The entities
      near the RP confirm its identity and validate names it claims.
      The fabric routes the message towards the appropriate IdP,
      validating the name of the IdP in the process.  The routing can be
      statically configured.  Alternatively, a routing protocol could be
      developed to exchange reachability information about a given IdP
      and to apply policy across the AAA fabric.  Such a routing
      protocol could flood naming constraints to the appropriate points
      in the fabric.








Howlett, et al.               Informational                    [Page 20]

RFC 7831                   ABFAB Architecture                   May 2016


   Trust Broker:
      Instead of routing messages through AAA proxies, some trust broker
      could establish keys between entities near the RP and entities
      near the IdP.  The advantage of this approach is efficiency of
      message handling.  Fewer entities are needed to be involved for
      each message.  Security may be improved by sending individual
      messages over fewer hops.  Rules determination involves decisions
      made by trust brokers about what keys to grant.  Also, associated
      with each credential is context about rules and about other
      aspects of technical trust, including names that may be claimed.
      A routing protocol similar to the one for AAA proxies is likely to
      be useful to trust brokers in flooding rules and naming
      constraints.

   Global Credential:
      A global credential such as a public key and certificate in a
      public key infrastructure can be used to establish technical
      trust.  A directory or distributed database such as the Domain
      Name System is used by the RP to discover the endpoint to contact
      for a given NAI.  Either the database or certificates can provide
      a place to store information about rules determination and naming
      constraints.  Provided that no intermediates are required (or
      appear to be required) and that the RP and IdP are sufficient to
      enforce and determine rules, rules determination is reasonably
      simple.  However, applying certain rules is likely to be quite
      complex.  For example, if multiple sets of rules are possible
      between an IdP and RP, confirming that the correct set is used may
      be difficult.  This is particularly true if intermediates are
      involved in making the decision.  Also, to the extent that
      directory information needs to be trusted, rules determination may
      be more complex.

   Real-world deployments are likely to be mixtures of these basic
   approaches.  For example, it will be quite common for an RP to route
   traffic to a AAA proxy within an organization.  That proxy could then
   use any of the above three methods to get closer to the IdP.  It is
   also likely that, rather than being directly reachable, the IdP may
   have a proxy on the edge of its organization.  Federations will
   likely provide a traditional AAA proxy interface even if they also
   provide another mechanism for increased efficiency or security.

2.1.4.  AAA Security



   For the AAA framework, there are two different places where security
   needs to be examined.  The first is the security that is in place for
   the links in the AAA backbone being used.  The second are the nodes
   that form the AAA backbone.




Howlett, et al.               Informational                    [Page 21]

RFC 7831                   ABFAB Architecture                   May 2016


   The default link security for RADIUS is showing its age, as it uses
   MD5 and a shared secret to both obfuscate passwords and provide
   integrity on the RADIUS messages.  While some EAP methods include the
   ability to protect the client authentication credentials, the MSK
   returned from the IdP to the RP is protected only by RADIUS security.
   In many environments, this is considered to be insufficient,
   especially as not all attributes are obfuscated and can thus leak
   information to a passive eavesdropper.  The use of RADIUS with
   Transport Layer Security (TLS) [RFC6614] and/or Datagram Transport
   Layer Security (DTLS) [RFC7360] addresses these attacks.  The same
   level of security is included in the base Diameter specifications.

2.1.5.  SAML Assertions



   For the traditional use of AAA frameworks, i.e., granting access to a
   network, an affirmative response from the IdP is sufficient.  In the
   ABFAB world, the RP may need to get significantly more additional
   information about the client before granting access.  ABFAB therefore
   has a requirement that it can transport an arbitrary set of
   attributes about the client from the IdP to the RP.

   The Security Assertion Markup Language (SAML)
   [OASIS.saml-core-2.0-os] was designed in order to carry an extensible
   set of attributes about a subject.  Since SAML is extensible in the
   attribute space, ABFAB has no immediate needs to update the core SAML
   specifications for our work.  It will be necessary to update IdPs
   that need to return SAML Assertions to RPs and for both the IdP and
   the RP to implement a new SAML profile designed to carry SAML
   Assertions in AAA.  The new profile can be found in [RFC7833].  As
   SAML statements will frequently be large, RADIUS servers and clients
   that deal with SAML statements will need to implement [RFC7499].

   There are several issues that need to be highlighted:

   o  The security of SAML Assertions.

   o  Namespaces and mapping of SAML attributes.

   o  Subject naming of entities.

   o  Making multiple queries about the subject(s).

   o  Level of assurance for authentication.

   SAML Assertions have an optional signature that can be used to
   protect and provide the origination of the assertion.  These
   signatures are normally based on asymmetric key operations and
   require that the verifier be able to check not only the cryptographic



Howlett, et al.               Informational                    [Page 22]

RFC 7831                   ABFAB Architecture                   May 2016


   operation but also the binding of the originator's name and the
   public key.  In a federated environment, it will not always be
   possible for the RP to validate the binding; for this reason, the
   technical trust established in the federation is used as an alternate
   method of validating the origination and integrity of the SAML
   Assertion.

   Attributes in a SAML Assertion are identified by a name string.  The
   name string is either assigned by the SAML issuer context or scoped
   by a namespace (for example, a URI or object identifier (OID)).  This
   means that the same attribute can have different name strings used to
   identify it.  In many cases, but not all, the federation agreements
   will determine what attributes and names can be used in a SAML
   statement.  This means that the RP needs to map from the SAML issuer
   or federation name, type, and semantic to the name, type, and
   semantics that the policies of the RP are written in.  In other
   cases, the Federation Substrate, in the form of proxies, will modify
   the SAML Assertions in transit to do the necessary name, type, and
   value mappings as the assertion crosses boundaries in the federation.
   If the proxies are modifying the SAML Assertion, then they will
   remove any signatures on the SAML Assertion, as changing the content
   of the SAML Assertion would invalidate the signature.  In this case,
   the technical trust is the required mechanism for validating the
   integrity of the assertion.  (The proxy could re-sign the SAML
   Assertion, but the same issues of establishing trust in the proxy
   would still exist.)  Finally, the attributes may still be in the
   namespace of the originating IdP.  When this occurs, the RP will need
   to get the required mapping operations from the federation agreements
   and do the appropriate mappings itself.

   [RFC7833] has defined a new SAML name format that corresponds to the
   NAI name form defined by [RFC7542].  This allows for easy name
   matching in many cases, as the name form in the SAML statement and
   the name form used in RADIUS or Diameter will be the same.  In
   addition to the NAI name form, [RFC7833] also defines a pair of
   implicit name forms corresponding to the client and the client's
   machine.  These implicit name forms are based on the Identity-Type
   enumeration defined in the Tunnel Extensible Authentication Protocol
   (TEAP) specification [RFC7170].  If the name form returned in a SAML
   statement is not based on the NAI, then it is a requirement on the
   EAP server that it validate that the subject of the SAML Assertion,
   if any, is equivalent to the subject identified by the NAI used in
   the RADIUS or Diameter session.

   RADIUS has the ability to deal with multiple SAML queries for those
   EAP servers that follow [RFC5080].  In this case, a State attribute
   will always be returned with the Access-Accept.  The EAP client can
   then send a new Access-Request with the State attribute and the new



Howlett, et al.               Informational                    [Page 23]

RFC 7831                   ABFAB Architecture                   May 2016


   SAML request.  Multiple SAML queries can then be done by making a new
   Access-Request, using the State attribute returned in the last
   Access-Accept to link together the different RADIUS sessions.

   Some RPs need to ensure that specific criteria are met during the
   authentication process.  This need is met by using levels of
   assurance.  A level of assurance is communicated to the RP from the
   EAP server by using a SAML Authentication Request, using the
   Authentication Profile described in [RFC7833].  When crossing
   boundaries between different federations, (1) the policy specified
   will need to be shared between the two federations, (2) the policy
   will need to be mapped by the proxy server on the boundary, or
   (3) the proxy server on the boundary will need to supply information
   to the EAP server so that the EAP server can do the required mapping.
   If this mapping is not done, then the EAP server will not be able to
   enforce the desired level of assurance, as it will not understand the
   policy requirements.

2.2.  Client to Identity Provider



   Looking at the communications between the client and the IdP, the
   following items need to be dealt with:

   o  The client and the IdP need to mutually authenticate each other.

   o  The client and the IdP need to mutually agree on the identity of
      the RP.

   ABFAB selected EAP for the purposes of mutual authentication and
   assisted in creating some new EAP channel-binding documents for
   dealing with determining the identity of the RP.  A framework for the
   channel-binding mechanism has been defined in [RFC6677] that allows
   the IdP to check the identity of the RP provided by the AAA framework
   against the identity provided by the client.

2.2.1.  Extensible Authentication Protocol (EAP)



   Traditional web federation does not describe how a client interacts
   with an IdP for authentication.  As a result, this communication is
   not standardized.  There are several disadvantages to this approach.
   Since the communication is not standardized, it is difficult for
   machines to recognize which entity is going to do the authentication,
   and thus which credentials to use and where in the authentication
   form the credentials are to be entered.  It is much easier for humans
   to correctly deal with these problems.  The use of browsers for
   authentication restricts the deployment of more secure forms of
   authentication beyond plaintext usernames and passwords known by the
   server.  In a number of cases, the authentication interface may be



Howlett, et al.               Informational                    [Page 24]

RFC 7831                   ABFAB Architecture                   May 2016


   presented before the client has adequately validated that they are
   talking to the intended server.  By giving control of the
   authentication interface to a potential attacker, the security of the
   system may be reduced, and opportunities for phishing may be
   introduced.

   As a result, it is desirable to choose some standardized approach for
   communication between the client's end host and the IdP.  There are a
   number of requirements this approach must meet, as noted below.

   Experience has taught us one key security and scalability
   requirement: it is important that the RP not get possession of the
   long-term secret of the client.  Aside from a valuable secret being
   exposed, a synchronization problem can develop when the client
   changes keys with the IdP.

   Since there is no single authentication mechanism that will be used
   everywhere, another associated requirement is that the authentication
   framework must allow for the flexible integration of authentication
   mechanisms.  For instance, some IdPs require hardware tokens, while
   others use passwords.  A service provider wants to provide support
   for both authentication methods and also for other methods from IdPs
   not yet seen.

   These requirements can be met by utilizing standardized and
   successfully deployed technology, namely the EAP framework [RFC3748].
   Figure 3 illustrates the integration graphically.

   EAP is an end-to-end framework; it provides for two-way communication
   between a peer (i.e., client or Individual) through the EAP
   authenticator (i.e., RP) to the back end (i.e., IdP).  This is
   precisely -- and conveniently -- the communication path that is
   needed for federated identity.  Although EAP support is already
   integrated in AAA systems (see [RFC3579] and [RFC4072]), several
   challenges remain:

   o  The first is how to carry EAP payloads from the end host to
      the RP.

   o  Another is to verify statements the RP has made to the client,
      confirm that these statements are consistent with statements made
      to the IdP, and confirm that all of the above are consistent with
      the federation and any federation-specific policy or
      configuration.

   o  Another challenge is choosing which IdP to use for which service.





Howlett, et al.               Informational                    [Page 25]

RFC 7831                   ABFAB Architecture                   May 2016


   The EAP method used for ABFAB needs to meet the following
   requirements:

   o  It needs to provide mutual authentication of the client and IdP.

   o  It needs to support channel binding.

   As of this writing, the only EAP method that meets these criteria is
   TEAP [RFC7170], either alone (if client certificates are used) or
   with an inner EAP method that does mutual authentication.

2.2.2.  EAP Channel Binding



   EAP channel binding is easily confused with a facility in GSS-API
   that is also called "channel binding".  GSS-API channel binding
   provides protection against man-in-the-middle attacks when GSS-API is
   used for authentication inside of some tunnel; it is similar to a
   facility called "cryptographic binding" in EAP.  See [RFC5056] for a
   discussion of the differences between these two facilities.

   The client knows, in theory, the name of the RP that it attempted to
   connect to; however, in the event that an attacker has intercepted
   the protocol, the client and the IdP need to be able to detect this
   situation.  A general overview of the problem, along with a
   recommended way to deal with the channel-binding issues, can be found
   in [RFC6677].

   Since the time that [RFC6677] was published, a number of possible
   attacks were found.  Methods to address these attacks have been
   outlined in [RFC7029].

2.3.  Client to Relying Party



   The final set of interactions between the parties to consider are
   those between the client and the RP.  In some ways, this is the most
   complex set, since at least part of it is outside the scope of the
   ABFAB work.  The interactions between these parties include:

   o  Running the protocol that implements the service that is provided
      by the RP and desired by the client.

   o  Authenticating the client to the RP and the RP to the client.

   o  Providing the necessary security services to the service protocol
      that it needs, beyond authentication.

   o  Dealing with client re-authentication where desired.




Howlett, et al.               Informational                    [Page 26]

RFC 7831                   ABFAB Architecture                   May 2016


2.3.1.  GSS-API



   One of the remaining layers is responsible for integration of
   federated authentication with the application.  Applications have
   adopted a number of approaches for providing security, so multiple
   strategies for integration of federated authentication with
   applications may be needed.  To this end, we start with a strategy
   that provides integration with a large number of application
   protocols.

   Many applications, such as Secure Shell (SSH) [RFC4462], NFS
   [RFC7530], DNS [RFC3645], and several non-IETF applications, support
   GSS-API [RFC2743].  Many applications, such as IMAP, SMTP, the
   Extensible Messaging and Presence Protocol (XMPP), and the
   Lightweight Directory Access Protocol (LDAP), support the Simple
   Authentication and Security Layer (SASL) [RFC4422] framework.  These
   two approaches work together nicely: by creating a GSS-API mechanism,
   SASL integration is also addressed.  In effect, using a GSS-API
   mechanism with SASL simply requires placing some headers before the
   mechanism's messages and constraining certain GSS-API options.

   GSS-API is specified in terms of an abstract set of operations that
   can be mapped into a programming language to form an API.  When
   people are first introduced to GSS-API, they focus on it as an API.
   However, from the perspective of authentication for non-web
   applications, GSS-API should be thought of as a protocol as well as
   an API.  When looked at as a protocol, it consists of abstract
   operations such as the initial context exchange, which includes two
   sub-operations (GSS_Init_sec_context and GSS_Accept_sec_context)
   [RFC2743].  An application defines which abstract operations it is
   going to use and where messages produced by these operations fit into
   the application architecture.  A GSS-API mechanism will define what
   actual protocol messages result from that abstract message for a
   given abstract operation.  So, since this work is focusing on a
   particular GSS-API mechanism, we generally focus on protocol elements
   rather than the API view of GSS-API.

   The API view of GSS-API does have significant value as well; since
   the abstract operations are well defined, the information that a
   mechanism gets from the application is well defined.  Also, the set
   of assumptions the application is permitted to make is generally well
   defined.  As a result, an application protocol that supports GSS-API
   or SASL is very likely to be usable with a new approach to
   authentication, including the authentication mechanism defined in
   this document, with no required modifications.  In some cases,
   support for a new authentication mechanism has been added using
   plugin interfaces to applications without the application being
   modified at all.  Even when modifications are required, they can



Howlett, et al.               Informational                    [Page 27]

RFC 7831                   ABFAB Architecture                   May 2016


   often be limited to supporting a new naming and authorization model.
   For example, this work focuses on privacy; an application that
   assumes that it will always obtain an identifier for the client will
   need to be modified to support anonymity, unlinkability, or
   pseudonymity.

   So, we use GSS-API and SASL because a number of the application
   protocols we wish to federate support these strategies for security
   integration.  What does this mean from a protocol standpoint, and how
   does this relate to other layers?  This means that we need to design
   a concrete GSS-API mechanism.  We have chosen to use a GSS-API
   mechanism that encapsulates EAP authentication.  So, GSS-API (and
   SASL) encapsulates EAP between the end host and the service.  The AAA
   framework encapsulates EAP between the RP and the IdP.  The GSS-API
   mechanism includes rules about how initiators and services are named
   as well as per-message security and other facilities required by the
   applications we wish to support.

2.3.2.  Protocol Transport



   The transport of data between the client and the RP is not provided
   by GSS-API.  GSS-API creates and consumes messages, but it does not
   provide the transport itself; instead, the protocol using GSS-API
   needs to provide the transport.  In many cases, HTTP or HTTPS is used
   for this transport, but other transports are perfectly acceptable.
   The core GSS-API document [RFC2743] provides some details on what
   requirements exist.

   In addition, we highlight the following:

   o  The transport does not need to provide either confidentiality or
      integrity.  After GSS-EAP has finished negotiation, GSS-API can be
      used to provide both services.  If the negotiation process itself
      needs protection from eavesdroppers, then the transport would need
      to provide the necessary services.

   o  The transport needs to provide reliable transport of the messages.

   o  The transport needs to ensure that tokens are delivered in order
      during the negotiation process.

   o  GSS-API messages need to be delivered atomically.  If the
      transport breaks up a message, it must also reassemble the message
      before delivery.







Howlett, et al.               Informational                    [Page 28]

RFC 7831                   ABFAB Architecture                   May 2016


2.3.3.  Re-authentication



   There are circumstances where the RP will want to have the client
   re-authenticate itself.  These include very long sessions, where the
   original authentication is time limited or cases where in order to
   complete an operation a different authentication is required.
   GSS-EAP does not have any mechanism for the server to initiate a
   re-authentication, as all authentication operations start from the
   client.  If a protocol using GSS-EAP needs to support
   re-authentication that is initiated by the server, then a request
   from the server to the client for the re-authentication to start
   needs to be placed in the protocol.

   Clients can reuse the existing secure connection established by
   GSS-API, and run the new authentication in that connection, by
   calling GSS_Init_sec_context.  At this point, a full
   re-authentication will be done.

3.  Application Security Services



   One of the key goals is to integrate federated authentication with
   existing application protocols and, where possible, existing
   implementations of these protocols.  Another goal is to perform this
   integration while meeting the best security practices of the
   technologies used to perform the integration.  This section describes
   security services and properties required by the EAP GSS-API
   mechanism in order to meet these goals.  This information could be
   viewed as specific to that mechanism.  However, other future
   application integration strategies are very likely to need similar
   services.  So, it is likely that these services will be expanded
   across application integration strategies if new application
   integration strategies are adopted.

3.1.  Authentication



   GSS-API provides an optional security service called "mutual
   authentication".  This service means that in addition to the
   initiator providing (potentially anonymous or pseudonymous) identity
   to the acceptor, the acceptor confirms its identity to the initiator.
   In the context of ABFAB in particular, the naming of this service is
   confusing.  We still say that mutual authentication is provided when
   the identity of an acceptor is strongly authenticated to an anonymous
   initiator.








Howlett, et al.               Informational                    [Page 29]

RFC 7831                   ABFAB Architecture                   May 2016


   Unfortunately, [RFC2743] does not explicitly talk about what mutual
   authentication means.  Within this document, we therefore define
   mutual authentication as follows:

   o  If a target name is configured for the initiator, then the
      initiator trusts that the supplied target name describes the
      acceptor.  This implies that (1) appropriate cryptographic
      exchanges took place for the initiator to make such a trust
      decision and (2) after evaluating the results of these exchanges,
      the initiator's policy trusts that the target name is accurate.

   o  If no target name is configured for the initiator, then the
      initiator trusts that the acceptor name, supplied by the acceptor,
      correctly names the entity it is communicating with.

   o  Both the initiator and acceptor have the same key material for
      per-message keys, and both parties have confirmed that they
      actually have the key material.  In EAP terms, there is a
      protected indication of success.

   Mutual authentication is an important defense against certain aspects
   of phishing.  Intuitively, clients would like to assume that if some
   party asks for their credentials as part of authentication,
   successfully gaining access to the resource means that they are
   talking to the expected party.  Without mutual authentication, the
   server could "grant access" regardless of what credentials are
   supplied.  Mutual authentication better matches this user intuition.

   It is important, therefore, that the GSS-EAP mechanism implement
   mutual authentication.  That is, an initiator needs to be able to
   request mutual authentication.  When mutual authentication is
   requested, only EAP methods capable of providing the necessary
   service can be used, and appropriate steps need to be taken to
   provide mutual authentication.  While a broader set of EAP methods
   could be supported by not requiring mutual authentication, it was
   decided that the client needs to always have the ability to request
   it.  In some cases, the IdP and the RP will not support mutual
   authentication; however, the client will always be able to detect
   this and make an appropriate security decision.

   The AAA infrastructure may hide the initiator's identity from the
   GSS-API acceptor, providing anonymity between the initiator and the
   acceptor.  At this time, whether the identity is disclosed is
   determined by EAP server policy rather than by an indication from the
   initiator.  Also, initiators are unlikely to be able to determine
   whether anonymous communication will be provided.  For this reason,
   initiators are unlikely to set the anonymous return flag from
   GSS_Init_sec_context (Section 2.2.1 of [RFC2743]).



Howlett, et al.               Informational                    [Page 30]

RFC 7831                   ABFAB Architecture                   May 2016


3.2.  GSS-API Channel Binding



   [RFC5056] defines a concept of channel binding that is used to
   prevent man-in-the-middle attacks.  This type of channel binding
   works by taking a cryptographic value from the transport security
   layer and checks to see that both sides of the GSS-API conversation
   know this value.  Transport Layer Security (TLS) [RFC5246] is the
   most common transport security layer used for this purpose.

   It needs to be stressed that channel binding as described in
   [RFC5056] (also called "GSS-API channel binding" when GSS-API is
   involved) is not the same thing as EAP channel binding.  GSS-API
   channel binding is used for detecting man-in-the-middle attacks.  EAP
   channel binding is used for mutual authentication and acceptor naming
   checks.  See [RFC7055] for details.  A more detailed description of
   the differences between the facilities can be found in [RFC5056].

   The use of TLS can provide both encryption and integrity on the
   channel.  It is common to provide SASL and GSS-API with these other
   security services.

   One of the benefits that the use of TLS provides is that a client has
   the ability to validate the name of the server.  However, this
   validation is predicated on a couple of things.  The TLS session
   needs to be using certificates and not be an anonymous session.  The
   client and the TLS server need to share a common trust point for the
   certificate used in validating the server.  TLS provides its own
   server authentication.  However, there are a variety of situations
   where, for policy or usability reasons, this authentication is not
   checked.  When the TLS authentication is checked, if the trust
   infrastructure behind the TLS authentication is different from the
   trust infrastructure behind the GSS-API mutual authentication, then
   confirming the endpoints using both trust infrastructures is likely
   to enhance security.  If the endpoints of the GSS-API authentication
   are different than the endpoints of the lower layer, this is a strong
   indication of a problem, such as a man-in-the-middle attack.  Channel
   binding provides a facility to determine whether these endpoints are
   the same.

   The GSS-EAP mechanism needs to support channel binding.  When an
   application provides channel-binding data, the mechanism needs to
   confirm that this is the same on both sides, consistent with the
   GSS-API specification.








Howlett, et al.               Informational                    [Page 31]

RFC 7831                   ABFAB Architecture                   May 2016


3.3.  Host-Based Service Names



   IETF security mechanisms typically take a host name and perhaps a
   service, entered by a user, and make some trust decision about
   whether the remote party in the interaction is the intended party.
   This decision can be made via the use of certificates, preconfigured
   key information, or a previous leap of trust.  GSS-API has defined a
   relatively flexible naming convention; however, most of the IETF
   applications that use GSS-API (including SSH, NFS, IMAP, LDAP, and
   XMPP) have chosen to use a more restricted naming convention based on
   the host name.  The GSS-EAP mechanism needs to support host-based
   service names in order to work with existing IETF protocols.

   The use of host-based service names leads to a challenging trust
   delegation problem.  Who is allowed to decide whether a particular
   host name maps to a specific entity?  Possible solutions to this
   problem have been looked at.

   o  The Public Key Infrastructure (PKI) used by the web has chosen to
      have a number of trust anchors (root certificate authorities),
      each of which can map any host name to a public key.

   o  A number of GSS-API mechanisms, such as Kerberos [RFC1964], have
      split the problem into two parts.  [RFC1964] introduced a new
      concept called a realm; the realm is responsible for host mapping
      within itself.  The mechanism then decides what realm is
      responsible for a given name.  This is the approach adopted by
      ABFAB.

   GSS-EAP defines a host naming convention that takes into account the
   host name, the realm, the service, and the service parameters.  An
   example of a GSS-API service name is "xmpp/foo@example.com".  This
   identifies the XMPP service on the host foo in the realm example.com.
   Any of the components, except for the service name, may be omitted
   from a name.  When omitted, a local default would be used for that
   component of the name.

   While there is no requirement that realm names map to Fully Qualified
   Domain Names (FQDNs) within DNS, in practice this is normally true.
   Doing so allows the realm portion of service names and the portion of
   NAIs to be the same.  It also allows for the use of DNS in locating
   the host of a service while establishing the transport channel
   between the client and the RP.

   It is the responsibility of the application to determine the server
   that it is going to communicate with; GSS-API has the ability to help
   confirm that the server is the desired server but not to determine
   the name of the server to use.  It is also the responsibility of the



Howlett, et al.               Informational                    [Page 32]

RFC 7831                   ABFAB Architecture                   May 2016


   application to determine how much of the information identifying the
   service needs to be validated by the ABFAB system.  The information
   that needs to be validated is used to construct the service name
   passed into the GSS-EAP mechanism.  What information is to be
   validated will depend on (1) what information was provided by the
   client and (2) what information is considered significant.  If the
   client only cares about getting a specific service, then it does not
   need to validate the host and realm that provides the service.

   Applications may retrieve information about providers of services
   from DNS.  Service Records (SRVs) [RFC2782] and Naming Authority
   Pointer (NAPTR) [RFC3401] records are used to help find a host that
   provides a service; however, the necessity of having DNSSEC on the
   queries depends on how the information is going to be used.  If the
   host name returned is not going to be validated by EAP channel
   binding because only the service is being validated, then DNSSEC
   [RFC4033] is not required.  However, if the host name is going to be
   validated by EAP channel binding, then DNSSEC needs to be used to
   ensure that the correct host name is validated.  In general, if the
   information that is returned from the DNS query is to be validated,
   then it needs to be obtained in a secure manner.

   Another issue that needs to be addressed for host-based service names
   is that they do not work ideally when different instances of a
   service are running on different ports.  If the services are
   equivalent, then it does not matter.  However, if there are
   substantial differences in the quality of the service, that
   information needs to be part of the validation process.  If one has
   just a host name and not a port in the information being validated,
   then this is not going to be a successful strategy.

3.4.  Additional GSS-API Services



   GSS-API provides per-message security services that can provide
   confidentiality and/or integrity.  Some IETF protocols, such as NFS
   and SSH, take advantage of these services.  As a result, GSS-EAP
   needs to support these services.  As with mutual authentication,
   per-message security services will limit the set of EAP methods that
   can be used to those that generate a Master Session Key (MSK).  Any
   EAP method that produces an MSK is able to support per-message
   security services as described in [RFC2743].

   GSS-API provides a pseudorandom function.  This function generates a
   pseudorandom sequence using the shared session key as the seed for
   the bytes generated.  This provides an algorithm that both the
   initiator and acceptor can run in order to arrive at the same key
   value.  The use of this feature allows an application to generate
   keys or other shared secrets for use in other places in the protocol.



Howlett, et al.               Informational                    [Page 33]

RFC 7831                   ABFAB Architecture                   May 2016


   In this regard, it is similar in concept to the mechanism (formerly
   known as "TLS Extractors") described in [RFC5705].  While no current
   IETF protocols require this feature, non-IETF protocols are expected
   to take advantage of it in the near future.  Additionally, a number
   of protocols have found the mechanism described in [RFC5705] to be
   useful in this regard, so it is highly probable that IETF protocols
   may also start using this feature.

4.  Privacy Considerations



   As an architecture designed to enable federated authentication and
   allow for the secure transmission of identity information between
   entities, ABFAB obviously requires careful consideration regarding
   privacy and the potential for privacy violations.

   This section examines the privacy-related information presented in
   this document, summarizing the entities that are involved in ABFAB
   communications and what exposure they have to identity information.
   In discussing these privacy considerations in this section, we use
   terminology and ideas from [RFC6973].

   Note that the ABFAB architecture uses at its core several existing
   technologies and protocols; detailed privacy discussion regarding
   these topics is not examined.  This section instead focuses on
   privacy considerations specifically related to the overall
   architecture and usage of ABFAB.

      +--------+       +---------------+       +--------------+
      | Client | <---> |      RP       | <---> | AAA Client   |
      +--------+       +---------------+       +--------------+
                                                     ^
                                                     |
                                                     v
                       +---------------+       +----------------+
                       | SAML Server   |       | AAA Proxy      |
                       +---------------+       | (or Proxies)   |
                                ^              +----------------+
                                |                       ^
                                |                       |
                                v                       v
      +------------+       +---------------+       +--------------+
      | EAP Server | <---> |   IdP         | <---> | AAA Server   |
      +------------+       +---------------+       +--------------+

                     Figure 4: Entities and Data Flow






Howlett, et al.               Informational                    [Page 34]

RFC 7831                   ABFAB Architecture                   May 2016


4.1.  Entities and Their Roles



   Categorizing the ABFAB entities shown in Figure 4 according to the
   taxonomy of terms from [RFC6973] is somewhat complicated, as the
   roles of each entity will change during the various phases of ABFAB
   communications.  The three main phases of relevance are the
   client-to-RP communication phase, the client-to-IdP (via the
   Federation Substrate) communication phase, and the IdP-to-RP (via the
   Federation Substrate) communication phase.

   In the client-to-RP communication phase, we have:

   Initiator:  Client.

   Observers:  Client, RP.

   Recipient:  RP.

   In the client-to-IdP (via the Federation Substrate) communication
   phase, we have:

   Initiator:  Client.

   Observers:  Client, RP, AAA Client, AAA Proxy (or Proxies), AAA
      Server, IdP.

   Recipient:  IdP

   In the IdP-to-RP (via the Federation Substrate) communication phase,
   we have:

   Initiator:  RP.

   Observers:  IdP, AAA Server, AAA Proxy (or Proxies), AAA Client, RP.

   Recipient:  IdP

   Eavesdroppers and attackers can reside on any or all communication
   links between the entities shown in Figure 4.












Howlett, et al.               Informational                    [Page 35]

RFC 7831                   ABFAB Architecture                   May 2016


   The various entities in the system might also collude or be coerced
   into colluding.  Some of the significant collusions to look at are as
   follows:

   o  If two RPs are colluding, they have the information available to
      both nodes.  This can be analyzed as if a single RP were offering
      multiple services.

   o  If an RP and a AAA proxy are colluding, then the trust of the
      system is broken, as the RP would be able to lie about its own
      identity to the IdP.  There is no known way to deal with this
      situation.

   o  If multiple AAA proxies are colluding, they can be treated as a
      single node for analysis.

   The Federation Substrate consists of all of the AAA entities.  In
   some cases, the AAA proxies may not exist, as the AAA client can talk
   directly to the AAA server.  Specifications such as the Trust Router
   Protocol (https://www.ietf.org/proceedings/86/slides/
   slides-86-rtgarea-0.pdf) and RADIUS dynamic discovery [RFC7585] can
   be used to shorten the path between the AAA client and the AAA server
   (and thus stop these AAA proxies from being observers); however, even
   in these circumstances, there may be AAA proxies in the path.

   In Figure 4, the IdP has been divided into multiple logical pieces;
   in actual implementations, these pieces will frequently be tightly
   coupled.  The links between these pieces provide the greatest
   opportunity for attackers and eavesdroppers to acquire information;
   however, as they are all under the control of a single entity, they
   are also the easiest to have tightly secured.

4.2.  Privacy Aspects of ABFAB Communication Flows



   In the ABFAB architecture, there are a few different types of data
   and identifiers in use.  The best way to understand them, and their
   potential privacy impacts, is to look at each phase of communication
   in ABFAB.

4.2.1.  Client to RP



   The flow of data between the client and the RP is divided into two
   parts.  The first part consists of all of the data exchanged as part
   of the ABFAB authentication process.  The second part consists of all
   of the data exchanged after the authentication process has been
   finished.





Howlett, et al.               Informational                    [Page 36]

RFC 7831                   ABFAB Architecture                   May 2016


   During the initial communication phase, the client sends an NAI (see
   [RFC7542]) to the RP.  Many EAP methods (but not all) allow the
   client to disclose an NAI to the RP in a form that includes only a
   realm component during this communication phase.  This is the minimum
   amount of identity information necessary for ABFAB to work -- it
   indicates an IdP that the principal has a relationship with.  EAP
   methods that do not allow this will necessarily also reveal an
   identifier for the principal in the IdP realm (e.g., a username).

   The data shared during the initial communication phase may be
   protected by a channel protocol such as TLS.  This will prevent the
   leakage of information to passive eavesdroppers; however, an active
   attacker may still be able to set itself up as a man-in-the-middle.
   The client may not be able to validate the certificates (if any)
   provided by the service, deferring the check of the identity of the
   RP until the completion of the ABFAB authentication protocol (using
   EAP channel binding rather than certificates).

   The data exchanged after the authentication process can have privacy
   and authentication using the GSS-API services.  If the overall
   application protocol allows for the process of re-authentication,
   then the same privacy implications as those discussed in previous
   paragraphs apply.

4.2.2.  Client to IdP (via Federation Substrate)



   This phase includes a secure TLS tunnel set up between the client and
   the IdP via the RP and Federation Substrate.  The process is
   initiated by the RP using the realm information given to it by the
   client.  Once set up, the tunnel is used to send credentials to the
   IdP to authenticate.

   Various operational information is transported between the RP and the
   IdP over the AAA infrastructure -- for example, using RADIUS headers.
   As no end-to-end security is provided by AAA, all AAA entities on the
   path between the RP and IdP have the ability to eavesdrop on this
   information.  Some of this information may form identifiers or
   explicit identity information:

   o  The RP knows the IP address of the client.  It is possible that
      the RP could choose to expose this IP address by including it in a
      RADIUS header (e.g., using the Calling-Station-Id).  This is a
      privacy consideration to take into account for the application
      protocol.

   o  The EAP MSK is transported between the IdP and the RP over the AAA
      infrastructure -- for example, through RADIUS headers.  This is a
      particularly important privacy consideration, as any AAA proxy



Howlett, et al.               Informational                    [Page 37]

RFC 7831                   ABFAB Architecture                   May 2016


      that has access to the EAP MSK is able to decrypt and eavesdrop on
      any traffic encrypted using that EAP MSK (i.e., all communications
      between the client and RP).  This problem can be mitigated if the
      application protocol sets up a secure tunnel between the client
      and the RP and performs a cryptographic binding between the tunnel
      and EAP MSK.

   o  Related to the bullet point above, the AAA server has access to
      the material necessary to derive the session key; thus, the AAA
      server can observe any traffic encrypted between the client and
      RP.  This "feature" was chosen as a simplification and to make
      performance faster; if it was decided that this trade-off was not
      desirable for privacy and security reasons, then extensions to
      ABFAB that make use of techniques such as Diffie-Hellman key
      exchange would mitigate this.

   The choice of EAP method used has other potential privacy
   implications.  For example, if the EAP method in use does not
   support mutual authentication, then there are no guarantees that the
   IdP is who it claims to be, and thus the full NAI, including a
   username and a realm, might be sent to any entity masquerading as a
   particular IdP.

   Note that ABFAB has not specified any AAA accounting requirements.
   Implementations that use the accounting portion of AAA should
   consider privacy appropriately when designing this aspect.

4.2.3.  IdP to RP (via Federation Substrate)



   In this phase, the IdP communicates with the RP, informing it as to
   the success or failure of authentication of the user and, optionally,
   the sending of identity information about the principal.

   As in the previous flow (client to IdP), various operation
   information is transported between the IdP and RP over the AAA
   infrastructure, and the same privacy considerations apply.  However,
   in this flow, explicit identity information about the authenticated
   principal can be sent from the IdP to the RP.  This information can
   be sent through RADIUS headers, or using SAML [RFC7833].  This can
   include protocol-specific identifiers, such as SAML NameIDs, as well
   as arbitrary attribute information about the principal.  What
   information will be released is controlled by policy on the IdP.  As
   before, when sending this information through RADIUS headers, all AAA
   entities on the path between the RP and IdP have the ability to
   eavesdrop, unless additional security measures are taken (such as the
   use of TLS for RADIUS [RFC6614]).  However, when sending this





Howlett, et al.               Informational                    [Page 38]

RFC 7831                   ABFAB Architecture                   May 2016


   information using SAML as specified in [RFC7833], confidentiality of
   the information should be guaranteed, as [RFC7833] requires the use
   of TLS for RADIUS.

4.3.  Relationship between User and Entities



   o  Between user and IdP - The IdP is an entity the user will have a
      direct relationship with, created when the organization that
      operates the entity provisioned and exchanged the user's
      credentials.  Privacy and data protection guarantees may form a
      part of this relationship.

   o  Between user and RP - The RP is an entity the user may or may not
      have a direct relationship with, depending on the service in
      question.  Some services may only be offered to those users where
      such a direct relationship exists (for particularly sensitive
      services, for example), while some may not require this and would
      instead be satisfied with basic federation trust guarantees
      between themselves and the IdP.  This may well include the option
      that the user stays anonymous with respect to the RP (though,
      obviously, never anonymous to the IdP).  If attempting to preserve
      privacy via data minimization (Section 1), then the only attribute
      information about Individuals exposed to the RP should be
      attribute information that is strictly necessary for the operation
      of the service.

   o  Between user and Federation Substrate - The user is highly likely
      to have no knowledge of, or relationship with, any entities
      involved with the Federation Substrate (not that the IdP and/or RP
      may, however).  Knowledge of attribute information about
      Individuals for these entities is not necessary, and thus such
      information should be protected in such a way as to prevent the
      possibility of access to this information.

4.4.  Accounting Information



   Alongside the core authentication and authorization that occur in AAA
   communications, accounting information about resource consumption may
   be delivered as part of the accounting exchange during the lifetime
   of the granted application session.

4.5.  Collection and Retention of Data and Identifiers



   In cases where RPs are not required to identify a particular
   Individual when an Individual wishes to make use of their service,
   the ABFAB architecture enables anonymous or pseudonymous access.
   Thus, data and identifiers other than pseudonyms and unlinkable
   attribute information need not be stored and retained.



Howlett, et al.               Informational                    [Page 39]

RFC 7831                   ABFAB Architecture                   May 2016


   However, in cases where RPs require the ability to identify a
   particular Individual (e.g., so they can link this identity
   information to a particular account in their service, or where
   identity information is required for audit purposes), the service
   will need to collect and store such information, and to retain it for
   as long as they require.  The de-provisioning of such accounts and
   information is out of scope for ABFAB, but for privacy protection, it
   is obvious that any identifiers collected should be deleted when they
   are no longer needed.

4.6.  User Participation



   In the ABFAB architecture, by its very nature users are active
   participants in the sharing of their identifiers, as they initiate
   the communications exchange every time they wish to access a server.
   They are, however, not involved in the control of information related
   to them that is transmitted from the IdP to the RP for authorization
   purposes; rather, this is under the control of policy on the IdP.
   Due to the nature of the AAA communication flows, with the current
   ABFAB architecture there is no place for a process of gaining user
   consent for the information to be released from the IdP to the RP.

5.  Security Considerations



   This document describes the architecture for Application Bridging for
   Federated Access Beyond web (ABFAB), and security is therefore the
   main focus.  Many of the items that are security considerations have
   already been discussed in Section 4 ("Privacy Considerations").
   Readers should be sure to read that section as well.

   There are many places in this document where TLS is used.  While in
   some places (e.g., client to RP) anonymous connections can be used,
   it is very important that TLS connections within the AAA
   infrastructure and between the client and the IdP be fully
   authenticated and, if using certificates, that revocation be checked
   as well.  When using anonymous connections between the client and the
   RP, all messages and data exchanged between those two entities will
   be visible to an active attacker.  In situations where the client is
   not yet on the network, the status_request extension [RFC6066] can be
   used to obtain revocation-checking data inside of the TLS protocol.
   Clients also need to get the trust anchor for the IdP configured
   correctly in order to prevent attacks; this is a difficult problem in
   general and is going to be even more difficult for kiosk
   environments.

   Selection of the EAP methods to be permitted by clients and IdPs is
   important.  The use of a tunneling method such as TEAP [RFC7170]
   allows other EAP methods to be used while hiding the contents of



Howlett, et al.               Informational                    [Page 40]

RFC 7831                   ABFAB Architecture                   May 2016


   those EAP exchanges from the RP and the AAA framework.  When
   considering inner EAP methods, the considerations outlined in
   [RFC7029] about binding the inner and outer EAP methods need to be
   taken into account.  Finally, one wants to have the ability to
   support channel binding in those cases where the client needs to
   validate that it is talking to the correct RP.

   In those places where SAML statements are used, RPs will generally be
   unable to validate signatures on the SAML statement, either because
   the signature has been stripped off by the IdP or because the RP is
   unable to validate the binding between the signer, the key used to
   sign, and the realm represented by the IdP.  For these reasons, it is
   required that IdPs do the necessary trust checking on the SAML
   statements and that RPs can trust the AAA infrastructure to keep the
   SAML statements valid.

   When a pseudonym is generated as a unique long-term identifier for a
   client by an IdP, care must be taken in the algorithm that it cannot
   easily be reverse-engineered by the service provider.  If it can be
   reverse-engineered, then the service provider can consult an oracle
   to determine if a given unique long-term identifier is associated
   with a different known identifier.

6.  References



6.1.  Normative References



   [RFC2743]  Linn, J., "Generic Security Service Application Program
              Interface Version 2, Update 1", RFC 2743,
              DOI 10.17487/RFC2743, January 2000,
              <http://www.rfc-editor.org/info/rfc2743>.

   [RFC2865]  Rigney, C., Willens, S., Rubens, A., and W. Simpson,
              "Remote Authentication Dial In User Service (RADIUS)",
              RFC 2865, DOI 10.17487/RFC2865, June 2000,
              <http://www.rfc-editor.org/info/rfc2865>.

   [RFC3579]  Aboba, B. and P. Calhoun, "RADIUS (Remote Authentication
              Dial In User Service) Support For Extensible
              Authentication Protocol (EAP)", RFC 3579,
              DOI 10.17487/RFC3579, September 2003,
              <http://www.rfc-editor.org/info/rfc3579>.

   [RFC3748]  Aboba, B., Blunk, L., Vollbrecht, J., Carlson, J., and H.
              Levkowetz, Ed., "Extensible Authentication Protocol
              (EAP)", RFC 3748, DOI 10.17487/RFC3748, June 2004,
              <http://www.rfc-editor.org/info/rfc3748>.




Howlett, et al.               Informational                    [Page 41]

RFC 7831                   ABFAB Architecture                   May 2016


   [RFC4072]  Eronen, P., Ed., Hiller, T., and G. Zorn, "Diameter
              Extensible Authentication Protocol (EAP) Application",
              RFC 4072, DOI 10.17487/RFC4072, August 2005,
              <http://www.rfc-editor.org/info/rfc4072>.

   [RFC6677]  Hartman, S., Ed., Clancy, T., and K. Hoeper, "Channel-
              Binding Support for Extensible Authentication Protocol
              (EAP) Methods", RFC 6677, DOI 10.17487/RFC6677, July 2012,
              <http://www.rfc-editor.org/info/rfc6677>.

   [RFC7055]  Hartman, S., Ed., and J. Howlett, "A GSS-API Mechanism for
              the Extensible Authentication Protocol", RFC 7055,
              DOI 10.17487/RFC7055, December 2013,
              <http://www.rfc-editor.org/info/rfc7055>.

   [RFC7542]  DeKok, A., "The Network Access Identifier", RFC 7542,
              DOI 10.17487/RFC7542, May 2015,
              <http://www.rfc-editor.org/info/rfc7542>.

   [RFC7833]  Howlett, J., Hartman, S., and A. Perez-Mendez, Ed., "A
              RADIUS Attribute, Binding, Profiles, Name Identifier
              Format, and Confirmation Methods for the Security
              Assertion Markup Language (SAML)", RFC 7833,
              DOI 10.17487/RFC7833, May 2016,
              <http://www.rfc-editor.org/info/rfc7833>.

6.2.  Informative References



   [NIST-SP.800-63-2]
              Burr, W., Dodson, D., Newton, E., Perlner, R., Polk, W.,
              Gupta, S., and E. Nabbus, "Electronic Authentication
              Guideline", NIST Special Publication 800-63-2,
              August 2013, <http://dx.doi.org/10.6028/NIST.SP.800-63-2>.

   [OASIS.saml-core-2.0-os]
              Cantor, S., Kemp, J., Philpott, R., and E. Maler,
              "Assertions and Protocols for the OASIS Security
              Assertion Markup Language (SAML) V2.0", OASIS
              Standard saml-core-2.0-os, March 2005,
              <http://docs.oasis-open.org/security/saml/v2.0/
              saml-core-2.0-os.pdf>.

   [RFC1964]  Linn, J., "The Kerberos Version 5 GSS-API Mechanism",
              RFC 1964, DOI 10.17487/RFC1964, June 1996,
              <http://www.rfc-editor.org/info/rfc1964>.






Howlett, et al.               Informational                    [Page 42]

RFC 7831                   ABFAB Architecture                   May 2016


   [RFC2782]  Gulbrandsen, A., Vixie, P., and L. Esibov, "A DNS RR for
              specifying the location of services (DNS SRV)", RFC 2782,
              DOI 10.17487/RFC2782, February 2000,
              <http://www.rfc-editor.org/info/rfc2782>.

   [RFC3401]  Mealling, M., "Dynamic Delegation Discovery System (DDDS)
              Part One: The Comprehensive DDDS", RFC 3401,
              DOI 10.17487/RFC3401, October 2002,
              <http://www.rfc-editor.org/info/rfc3401>.

   [RFC3645]  Kwan, S., Garg, P., Gilroy, J., Esibov, L., Westhead, J.,
              and R. Hall, "Generic Security Service Algorithm for
              Secret Key Transaction Authentication for DNS (GSS-TSIG)",
              RFC 3645, DOI 10.17487/RFC3645, October 2003,
              <http://www.rfc-editor.org/info/rfc3645>.

   [RFC4033]  Arends, R., Austein, R., Larson, M., Massey, D., and S.
              Rose, "DNS Security Introduction and Requirements",
              RFC 4033, DOI 10.17487/RFC4033, March 2005,
              <http://www.rfc-editor.org/info/rfc4033>.

   [RFC4422]  Melnikov, A., Ed., and K. Zeilenga, Ed., "Simple
              Authentication and Security Layer (SASL)", RFC 4422,
              DOI 10.17487/RFC4422, June 2006,
              <http://www.rfc-editor.org/info/rfc4422>.

   [RFC4462]  Hutzelman, J., Salowey, J., Galbraith, J., and V. Welch,
              "Generic Security Service Application Program Interface
              (GSS-API) Authentication and Key Exchange for the Secure
              Shell (SSH) Protocol", RFC 4462, DOI 10.17487/RFC4462,
              May 2006, <http://www.rfc-editor.org/info/rfc4462>.

   [RFC5056]  Williams, N., "On the Use of Channel Bindings to Secure
              Channels", RFC 5056, DOI 10.17487/RFC5056, November 2007,
              <http://www.rfc-editor.org/info/rfc5056>.

   [RFC5080]  Nelson, D. and A. DeKok, "Common Remote Authentication
              Dial In User Service (RADIUS) Implementation Issues and
              Suggested Fixes", RFC 5080, DOI 10.17487/RFC5080,
              December 2007, <http://www.rfc-editor.org/info/rfc5080>.

   [RFC5246]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.2", RFC 5246,
              DOI 10.17487/RFC5246, August 2008,
              <http://www.rfc-editor.org/info/rfc5246>.






Howlett, et al.               Informational                    [Page 43]

RFC 7831                   ABFAB Architecture                   May 2016


   [RFC5705]  Rescorla, E., "Keying Material Exporters for Transport
              Layer Security (TLS)", RFC 5705, DOI 10.17487/RFC5705,
              March 2010, <http://www.rfc-editor.org/info/rfc5705>.

   [RFC5801]  Josefsson, S. and N. Williams, "Using Generic Security
              Service Application Program Interface (GSS-API) Mechanisms
              in Simple Authentication and Security Layer (SASL): The
              GS2 Mechanism Family", RFC 5801, DOI 10.17487/RFC5801,
              July 2010, <http://www.rfc-editor.org/info/rfc5801>.

   [RFC6066]  Eastlake 3rd, D., "Transport Layer Security (TLS)
              Extensions: Extension Definitions", RFC 6066,
              DOI 10.17487/RFC6066, January 2011,
              <http://www.rfc-editor.org/info/rfc6066>.

   [RFC6614]  Winter, S., McCauley, M., Venaas, S., and K. Wierenga,
              "Transport Layer Security (TLS) Encryption for RADIUS",
              RFC 6614, DOI 10.17487/RFC6614, May 2012,
              <http://www.rfc-editor.org/info/rfc6614>.

   [RFC6733]  Fajardo, V., Ed., Arkko, J., Loughney, J., and G. Zorn,
              Ed., "Diameter Base Protocol", RFC 6733,
              DOI 10.17487/RFC6733, October 2012,
              <http://www.rfc-editor.org/info/rfc6733>.

   [RFC6749]  Hardt, D., Ed., "The OAuth 2.0 Authorization Framework",
              RFC 6749, DOI 10.17487/RFC6749, October 2012,
              <http://www.rfc-editor.org/info/rfc6749>.

   [RFC6973]  Cooper, A., Tschofenig, H., Aboba, B., Peterson, J.,
              Morris, J., Hansen, M., and R. Smith, "Privacy
              Considerations for Internet Protocols", RFC 6973,
              DOI 10.17487/RFC6973, July 2013,
              <http://www.rfc-editor.org/info/rfc6973>.

   [RFC7029]  Hartman, S., Wasserman, M., and D. Zhang, "Extensible
              Authentication Protocol (EAP) Mutual Cryptographic
              Binding", RFC 7029, DOI 10.17487/RFC7029, October 2013,
              <http://www.rfc-editor.org/info/rfc7029>.

   [RFC7170]  Zhou, H., Cam-Winget, N., Salowey, J., and S. Hanna,
              "Tunnel Extensible Authentication Protocol (TEAP)
              Version 1", RFC 7170, DOI 10.17487/RFC7170, May 2014,
              <http://www.rfc-editor.org/info/rfc7170>.







Howlett, et al.               Informational                    [Page 44]

RFC 7831                   ABFAB Architecture                   May 2016


   [RFC7360]  DeKok, A., "Datagram Transport Layer Security (DTLS) as a
              Transport Layer for RADIUS", RFC 7360,
              DOI 10.17487/RFC7360, September 2014,
              <http://www.rfc-editor.org/info/rfc7360>.

   [RFC7499]  Perez-Mendez, A., Ed., Marin-Lopez, R., Pereniguez-Garcia,
              F., Lopez-Millan, G., Lopez, D., and A. DeKok, "Support of
              Fragmentation of RADIUS Packets", RFC 7499,
              DOI 10.17487/RFC7499, April 2015,
              <http://www.rfc-editor.org/info/rfc7499>.

   [RFC7530]  Haynes, T., Ed., and D. Noveck, Ed., "Network File System
              (NFS) Version 4 Protocol", RFC 7530, DOI 10.17487/RFC7530,
              March 2015, <http://www.rfc-editor.org/info/rfc7530>.

   [RFC7585]  Winter, S. and M. McCauley, "Dynamic Peer Discovery for
              RADIUS/TLS and RADIUS/DTLS Based on the Network Access
              Identifier (NAI)", RFC 7585, DOI 10.17487/RFC7585,
              October 2015, <http://www.rfc-editor.org/info/rfc7585>.

   [WS-TRUST] Lawrence, K., Kaler, C., Nadalin, A., Goodner, M., Gudgin,
              M., Turner, D., Barbir, A., and H. Granqvist,
              "WS-Trust 1.4", OASIS Standard ws-trust-2012-04,
              April 2012, <http://docs.oasis-open.org/ws-sx/ws-trust/
              v1.4/ws-trust.html>.


























Howlett, et al.               Informational                    [Page 45]

RFC 7831                   ABFAB Architecture                   May 2016


Acknowledgments

   We would like to thank Mayutan Arumaithurai, Klaas Wierenga, and Rhys
   Smith for their feedback.  Additionally, we would like to thank Eve
   Maler, Nicolas Williams, Bob Morgan, Scott Cantor, Jim Fenton, Paul
   Leach, and Luke Howard for their feedback on the federation
   terminology question.

   Furthermore, we would like to thank Klaas Wierenga for his review of
   the first draft version of this document.  We also thank Eliot Lear
   for his work on early draft versions of this document.

Authors' Addresses



   Josh Howlett
   Jisc
   Lumen House, Library Avenue, Harwell
   Oxford  OX11 0SG
   United Kingdom

   Phone: +44 1235 822363
   Email: Josh.Howlett@ja.net


   Sam Hartman
   Painless Security

   Email: hartmans-ietf@mit.edu


   Hannes Tschofenig
   ARM Ltd.
   110 Fulbourn Road
   Cambridge  CB1 9NJ
   United Kingdom

   Email: Hannes.tschofenig@gmx.net
   URI:   http://www.tschofenig.priv.at


   Jim Schaad
   August Cellars

   Email: ietf@augustcellars.com







Howlett, et al.               Informational                    [Page 46]