Network Working Group                                       J. Hutzelman
Internet-Draft                                                       CMU
Expires: July 15, 2002 January 3, 2003                                      J. Salowey
                                                           Cisco Systems
                                                            J. Galbraith
                                             Van Dyke Technologies, Inc.
                                                                V. Welch
                                                         U Chicago / ANL
                                                        January 14,
                                                            July 5, 2002

  GSSAPI Authentication and Key Exchange for the Secure Shell Protocol
                      draft-ietf-secsh-gsskeyex-03
                      draft-ietf-secsh-gsskeyex-04

Status of this Memo

   This document is an Internet-Draft and is in full conformance with
   all provisions of Section 10 of RFC2026.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF), its areas, and its working groups. Note that
   other groups may also distribute working documents as
   Internet-Drafts.

   Internet-Drafts are draft documents valid for a maximum of six
   months and may be updated, replaced, or obsoleted by other documents
   at any time. It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   The list of current Internet-Drafts can be accessed at
   http://www.ietf.org/ietf/1id-abstracts.txt.

   The list of Internet-Draft Shadow Directories can be accessed at
   http://www.ietf.org/shadow.html.

   This Internet-Draft will expire on July 15, 2002. January 3, 2003.

Copyright Notice

   Copyright (C) The Internet Society (2002). All Rights Reserved.

Abstract

   The Secure Shell protocol (SSH) is a protocol for secure remote
   login and other secure network services over an insecure network.

   The Generic Security Service Application Program Interface (GSS-API)
   [2] provides security services to callers in a mechanism-independent
   fashion.

   This memo describes methods for using the GSS-API for authentication
   and key exchange in SSH. It defines an SSH user authentication
   method which uses a specified GSSAPI mechanism to authenticate a
   user, and a family of SSH key exchange methods which use GSSAPI to
   authenticate the Diffie-Hellman exchange described in [11].

   This memo also defines a new host public key algorithm which can be
   used when no operations are needed using a host's public key, and a
   new user authentication method which allows an authorization name to
   be used in conjunction with any authentication which has already
   occurred as a side-effect of key exchange.

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [7].

1. Introduction

   This document describes the methods used to perform key exchange and
   user authentication in the Secure Shell protocol using the GSSAPI.
   To do this, it defines a family of key exchange methods, two user
   authentication methods, and a new host key algorithm. These
   definitions allow any GSSAPI mechanism to be used with the Secure
   Shell protocol.

   This document should be read only after reading the documents
   describing the SSH protocol architecture [9], transport layer
   protocol [11], and user authentication protocol [12].  This document
   freely uses terminology and notation from the architecture document
   without reference or further explanation.

1.1 SSH terminology

   The data types used in the packets are defined in the SSH
   architecture document [9].  It is particularly important to note the
   definition of string allows binary content.

   The SSH_MSG_USERAUTH_REQUEST packet refers to a service; this
   service name is an SSH service name, and has no relationship to
   GSSAPI service names.  Currently, the only defined service name is
   "ssh-connection", which refers to the SSH connection protocol [10].

2. GSSAPI Authenticated Diffie-Hellman Key Exchange

   This section defines a class of key exchange methods which combine
   the Diffie-Hellman key exchange from section 6 of [11] with mutual
   authentication using GSSAPI.

   Since the GSSAPI key exchange methods described in this section do
   not require the use of public key signature or encryption
   algorithms, they MAY be used with any host key algorithm, including
   the "null" algorithm described in Section 5.

2.1 Generic GSSAPI Key Exchange

   The following symbols are used in this description:

   o  C is the client, and S is the server

   o  p is a large safe prime, g is a generator for a subgroup of
      GF(p), and q is the order of the subgroup

   o  V_S is S's version string, and V_C is C's version string

   o  I_C is C's KEXINIT message, and I_S is S's KEXINIT message

   1.  C generates a random number x (1 < x < q) and computes e = g^x
       mod p.

   2.  C calls GSS_Init_sec_context, using the most recent reply token
       received from S during this exchange, if any.  For this call,
       the client MUST set the mutual_req_flag to "true" to request
       that mutual authentication be performed.  It also MUST set the
       integ_req_flag to "true" to request that per-message integrity
       protection be supported for this context.  In addition, the
       deleg_req_flag MAY be set to "true" to request access
       delegation, if requested by the user. Since the key exchange
       process authenticates only the host, the setting of the
       anon_req_flag is immaterial to this process.  If the client does
       not support the "external-keyx" user authentication method
       described in Section 4, or does not intend to use that method,
       then the anon_req_flag SHOULD be set to "true". Otherwise, this
       flag MAY be set to true if the client wishes to hide its
       identity.

       *  If the resulting major_status code is GSS_S_COMPLETE and the
          mutual_state flag is not true, then mutual authentication has
          not been established, and the key exchange MUST fail.

       *  If the resulting major_status code is GSS_S_COMPLETE and the
          integ_avail flag is not true, then per-message integrity
          protection is not available, and the key exchange MUST fail.

       *  If the resulting major_status code is GSS_S_COMPLETE and both
          the mutual_state and integ_avail flags are true, the
          resulting output token is sent to S.

       *  If the resulting major_status code is GSS_S_CONTINUE_NEEDED,
          the the output_token is sent to S, which will reply with a
          new token to be provided to GSS_Init_sec_context.

       *  The client MUST also include "e" with the first message it
          sends to the server during this process; if the server
          receives more than one "e" or none at all, the key exchange
          fails.

       *  It is an error if the call does not produce a token of
          non-zero length to be sent to the server.  In this case, the
          key exchange MUST fail.

   3.  S calls GSS_Accept_sec_context, using the token received from C.

       *  If the resulting major_status code is GSS_S_COMPLETE and the
          mutual_state flag is not true, then mutual authentication has
          not been established, and the key exchange MUST fail.

       *  If the resulting major_status code is GSS_S_COMPLETE and the
          integ_avail flag is not true, then per-message integrity
          protection is not available, and the key exchange MUST fail.

       *  If the resulting major_status code is GSS_S_COMPLETE and both
          the mutual_state and integ_avail flags are true, then the
          security context has been established, and processing
          continues with step 4.

       *  If the resulting major_status code is GSS_S_CONTINUE_NEEDED,
          then the output token is sent to C, and processing continues
          with step 2.

       *  If the resulting major_status code is GSS_S_COMPLETE, but a
          non-zero-length reply token is returned, then that token is
          sent to the client.

   4.  S generates a random number y (0 < y < q) and computes f = g^y
       mod p.  It computes K = e ^ y mod p, and H = hash(V_C || V_S ||
       I_C || I_S || K_S || e || f || K).  It then calls GSS_GetMIC to
       obtain a GSSAPI message integrity code for H.  S then sends f
       and the MIC to C.

   5.  This step is performed only if the server's final call to
       GSS_Accept_sec_context produced a non-zero-length final reply
       token to be sent to the client _and_ no previous call by the
       client to GSS_Init_sec_context has resulted in a major_status of
       GSS_S_COMPLETE.  Under these conditions, the client makes an
       additional call to GSS_Init_sec_context to process the final
       reply token.  This call is made exactly as described above.
       However, if the resulting major_status is anything other than
       GSS_S_COMPLETE, or a non-zero-length token is returned, it is an
       error and the key exchange MUST fail.

   6.  C computes K = f^x mod p, and H = hash(V_C || V_S || I_C || I_S
       || K_S || e || f || K).  It then calls GSS_VerifyMIC to verify
       that the MIC sent by S matches H.

   Either side MUST NOT send or accept e or f values that are not in
   the range [1, p-1].  If this condition is violated, the key exchange
   fails.

   If any call to GSS_Init_sec_context or GSS_Accept_sec_context
   returns a major_status other than GSS_S_COMPLETE or
   GSS_S_CONTINUE_NEEDED, or any other GSSAPI call returns a
   major_status other than GSS_S_COMPLETE, the key exchange fails.  If
   the key exchange fails due to a GSSAPI error on the server, the
   server SHOULD send a message informing the client of the details of
   the error before terminating the connection as required by [11].

   This is implemented with the following messages.  The hash algorithm
   for computing the exchange hash is defined by the method name, and
   is called HASH.  The group used for Diffie-Hellman key exchange and
   the underlying GSSAPI mechanism are also defined by the method name.

   After the client's first call to GSS_Init_sec_context, it sends the
   following:

           byte      SSH_MSG_KEXGSS_INIT
           string    output_token (from GSS_Init_sec_context)
           mpint     e

   Upon receiving the SSH_MSG_KEXGSS_INIT message, the server MAY send
   the following message, prior to any other messages, to inform the
   client of its host key.

           byte      SSH_MSG_KEXGSS_HOSTKEY
           string    server public host key and certificates (K_S)

   Since this key exchange method does not require the host key to be
   used for any encryption operations, this message is OPTIONAL.  If
   the "null" host key algorithm described in Section 5 is used, this
   message MUST NOT be sent.  If this message is sent, the server
   public host key(s) and/or certificate(s) in this message are encoded
   as a single string, in the format specified by the public key type
   in use (see [11], section 4.6).

   Each time the server's call to GSS_Accept_sec_context returns a
   major_status code of GSS_S_CONTINUE_NEEDED, it sends the following
   reply to the client:

           byte      SSH_MSG_KEXGSS_CONTINUE
           string    output_token (from GSS_Accept_sec_context)

   If the client receives this message after a call to
   GSS_Init_sec_context has returned a major_status code of
   GSS_S_COMPLETE, a protocol error has occurred and the key exchange
   MUST fail.

   Each time the client receives the message described above, it makes
   another call to GSS_Init_sec_context.  It then sends the following:

           byte      SSH_MSG_KEXGSS_CONTINUE
           string    output_token (from GSS_Init_sec_context)

   The server and client continue to trade these two messages as long
   as the server's calls to GSS_Accept_sec_context result in
   major_status codes of GSS_S_CONTINUE_NEEDED.  When a call results in
   a major_status code of GSS_S_COMPLETE, it sends one of two final
   messages.

   If the server's final call to GSS_Accept_sec_context (resulting in a
   major_status code of GSS_S_COMPLETE) returns a non-zero-length token
   to be sent to the client, it sends the following:

           byte      SSH_MSG_KEXGSS_COMPLETE
           mpint     f
           string    per_msg_token (MIC of H)
           boolean   TRUE
           string    output_token (from GSS_Accept_sec_context)

   If the client receives this message after a call to
   GSS_Init_sec_context has returned a major_status code of
   GSS_S_COMPLETE, a protocol error has occurred and the key exchange
   MUST fail.

   If the server's final call to GSS_Accept_sec_context (resulting in a
   major_status code of GSS_S_COMPLETE) returns a zero-length token or
   no token at all, it sends the following:

           byte      SSH_MSG_KEXGSS_COMPLETE
           mpint     f
           string    per_msg_token (MIC of H)
           boolean   FALSE

   If the client receives this message when no call to
   GSS_Init_sec_context has yet resulted in a major_status code of
   GSS_S_COMPLETE, a protocol error has occurred and the key exchange
   MUST fail.

   In the event of a GSSAPI error on the server, the server may send
   the following message before terminating the connection:

           byte      SSH_MSG_KEXGSS_ERROR
           uint32    major_status
           uint32    minor_status
           string    message
           string    language tag

   The message text MUST be encoded in the UTF-8 encoding described in
   [13].  Language tags are those described in [14].  Note that the
   message text may contain multiple lines separated by carriage
   return-line feed (CRLF) sequences. Application developers should
   take this into account when displaying these messages.

   The hash H is computed as the HASH hash of the concatenation of the
   following:

           string    V_C, the client's version string (CR and NL excluded)
           string    V_S, the server's version string (CR and NL excluded)
           string    I_C, the payload of the client's SSH_MSG_KEXINIT
           string    I_S, the payload of the server's SSH_MSG_KEXINIT
           string    K_S, the host key
           mpint     e, exchange value sent by the client
           mpint     f, exchange value sent by the server
           mpint     K, the shared secret

   This value is called the exchange hash, and it is used to
   authenticate the key exchange.  The exchange hash SHOULD be kept
   secret.  If no SSH_MSG_KEXGSS_HOSTKEY message has been send sent by the
   client
   server or received by the server, client, then the empty string is used in
   place of K_S when computing the exchange hash.

   The GSS_GetMIC call MUST be applied over H, not the original data.

2.2 gss-group1-sha1-*

   Each of these methods specifies GSSAPI authenticated Diffie-Hellman
   key exchange as described in Section 2.1 with SHA-1 as HASH, and the
   group defined in section 6.1 of [11].  The method name for each
   method is the concatenation of the string "gss-group1-sha1-" with
   the Base64 encoding of the MD5 hash [5] of the ASN.1 DER encoding
   [1] of the underlying GSSAPI mechanism's OID.  Base64 encoding is
   described in section 6.8 of [6].

   Each and every such key exchange method is implicitly registered by
   this specification.  The IESG is considered to be the owner of all
   such key exchange methods; this does NOT imply that the IESG is
   considered to be the owner of the underlying GSSAPI mechanism.

2.3 Other GSSAPI key exchange methods

   Key exchange method names starting with "gss-" are reserved for key
   exchange methods which conform to this document; in particular, for
   those methods which use the GSSAPI authenticated Diffie-Hellman key
   exchange algorithm described in Section 2.1, including any future
   methods which use different groups and/or hash functions.  The
   intent is that the names for any such future methods methods be
   defined in a similar manner to that used in Section 2.2.

3. GSSAPI User Authentication

   This section describes a general-purpose user authentication method
   based on [2].  It is intended to be run over the SSH user
   authentication protocol [12].

   The authentication method name for this protocol is "gssapi".

3.1 GSSAPI Authentication Overview

   GSSAPI authentication must maintain a context.  Authentication
   begins when the client sends a SSH_MSG_USERAUTH_REQUEST, which
   specifies the mechanism OIDs the client supports.

   If the server supports any of the requested mechanism OIDs, the
   server sends a SSH_MSG_USERAUTH_GSSAPI_RESPONSE message containing
   the mechanism OID.

   After the client receives SSH_MSG_USERAUTH_GSSAPI_RESPONSE, the
   client and server exchange SSH_MSG_USERAUTH_GSSAPI_TOKEN packets
   until the authentication mechanism either succeeds or fails.

   If at any time during the exchange, the client sends a new
   SSH_MSG_USERAUTH_REQUEST packet, the GSSAPI context is completely
   discarded and destroyed, and any further GSSAPI authentication MUST
   restart from the beginning.

3.2 Initiating GSSAPI authentication

   The GSSAPI authentication method is initiated when the client sends
   a SSH_MSG_USERAUTH_REQUEST:

           byte      SSH_MSG_USERAUTH_REQUEST
           string    user name (in ISO-10646 UTF-8 encoding)
           string    service name (in US-ASCII)
           string    "gssapi" (US-ASCII method name)
           uint32    n, the number of mechanism OIDs client supports
           string[n] mechanism OIDs

   Mechanism OIDs are encoded according to the ASN.1 basic encoding
   rules (BER), as described in [1] and in section 3.1 of [2].  The
   mechanism OIDs MUST be listed in order of preference, and the server
   must choose the first mechanism OID on the list that it supports.

   The client SHOULD NOT send more then one gssapi mechanism OID unless
   there are no non-GSSAPI authentication methods between the GSSAPI
   mechanisms in the order of preference, otherwise, authentication
   methods may be executed out of order.

   If the server does not support any of the specified OIDs, the server
   MUST fail the request by sending a SSH_MSG_USERAUTH_FAILURE packet.

   The user name may be an empty string if it can be deduced from the
   results of the gssapi authentication.  If the user name is not
   empty, and the requested user does not exist, the server MAY
   disconnect, or MAY send a bogus list of acceptable authentications
   but never accept any.  This makes it possible for the server to
   avoid disclosing information about which accounts exist.  In any
   case, if the user does not exist, the authentication request MUST
   NOT be accepted.

   The client MAY at any time continue with a new
   SSH_MSG_USERAUTH_REQUEST message, in which case the server MUST
   abandon the previous authentication attempt and continue with the
   new one.

3.3 Initial server response

   The server responds to the SSH_MSG_USERAUTH_REQUEST with either a
   SSH_MSG_USERAUTH_FAILURE if none of the mechanisms are supported, or
   with SSH_MSG_USERAUTH_GSSAPI_RESPONSE as follows:

           byte        SSH_MSG_USERAUTH_GSSAPI_RESPONSE
           string      selected mechanism OID

   The mechanism OID must be one of the OIDs sent by the client in the
   SSH_MSG_USERAUTH_REQUEST packet.

3.4 GSSAPI session

   Once the mechanism OID has been selected, the client will then
   initiate an exchange of one or more pairs of
   SSH_MSG_USERAUTH_GSSAPI_TOKEN packets.  These packets contain the
   tokens produced from the 'GSS_Init_sec_context()' and
   'GSS_Accept_sec_context()' calls.  The actual number of packets
   exchanged is determined by the underlying GSSAPI mechanism.

           byte        SSH_MSG_USERAUTH_GSSAPI_TOKEN
           string      data returned from either GSS_Init_sec_context()
                       or GSS_Accept_sec_context()

   If an error occurs during this exchange on server side, the server
   can terminate the method by sending a SSH_MSG_USERAUTH_FAILURE
   packet.  If an error occurs on client side, the client can terminate
   the method by sending a new SSH_MSG_USERAUTH_REQUEST packet.

   The client MAY use the deleg_req_flag in calls to
   GSS_Init_sec_context() to request credential delegation.

   Additional SSH_MSG_USERAUTH_GSSAPI_TOKEN messages are sent if and
   only if the calls to the GSSAPI routines produce send tokens of
   non-zero length.

   Any major status code other than GSS_S_COMPLETE or
   GSS_S_CONTINUE_NEEDED SHOULD be a failure.

3.5 Client acknowledgement

   It is possible for the server to successfully complete the GSSAPI
   method and the client to fail.  If the server simply assumed success
   on the part of the client and completed the authentication service,
   it is possible that the client would fail to complete the
   authentication method, but not be able to retry other methods
   because the server had already moved on.

   Therefore, the client MUST send the following message when it has
   successfully called GSS_Init_sec_context() and gotten GSS_S_COMPLETE:

           byte      SSH_MSG_USERAUTH_GSSAPI_EXCHANGE_COMPLETE

   This message MUST be sent if and only if GSS_Init_sec_context()
   returned GSS_S_COMPLETE.  If a token is returned then the
   SSH_MSG_USERAUTH_GSSAPI_TOKEN message MUST be sent before this one.

   If GSS_Init_sec_context() failed, the client MUST terminate the
   method by sending a new SSH_MSG_USERAUTH_REQUEST.

3.6 Completion

   As with all SSH authentication methods, successful completion is
   indicated by a SSH_MSG_USERAUTH_SUCCESS if no other authentication
   is required, or a SSH_MSG_USERAUTH_FAILURE with the partial success
   flag set if the server requires further authentication.

   This packet should be sent immediately following receipt of the the
   SSH_MSG_USERAUTH_GSSAPI_EXCHANGE_COMPLETE packet.

3.7 Error Status

   In the event a GSSAPI error occurs on the server during context
   establishment, the server SHOULD send the following message to
   inform the client of the details of the error before sending a
   SSH_MSG_USERAUTH_FAILURE message:

           byte      SSH_MSG_USERAUTH_GSSAPI_ERROR
           uint32    major_status
           uint32    minor_status
           string    message
           string    language tag

   The message text MUST be encoded in the UTF-8 encoding described in
   [13].  Language tags are those described in [14].  Note that the
   message text may contain multiple lines separated by carriage
   return-line feed (CRLF) sequences. Application developers should
   take this into account when displaying these messages.

   Clients receiving this message MAY log the error details and/or
   report them to the user.  Any server sending this message MUST
   ignore any SSH_MSG_UNIMPLEMENTED sent by the client in response.

4. External Key Exchange User Authentication

   This section describes a user authentication method building on the
   framework described in [12].  This method relies upon the key
   exchange to authenticate both the client and the server.  If the key
   exchange did not successfully perform these functions then the
   server MUST always respond to this request with
   SSH_MSG_USERAUTH_FAILURE with partial success set to false.

   The new mechanism is defined as follows:

         byte      SSH_MSG_USERAUTH_REQUEST
         string    user name (in ISO-10646 UTF-8 encoding)
         string    service name (in US-ASCII)
         string    "external-keyx" (US-ASCII method name)

   If the authentication performed as part of key exchange can be used
   to authorize login as the requested user, this method is successful,
   and the server responds with SSH_MSG_USERAUTH_SUCCESS if no more
   authentications are needed, or with SSH_MSG_USERAUTH_FAILURE with
   partial success set to true if more authentications are needed.

   If the authentication performed as part of key-exchange cannot be
   used to authorize login as the requested user, then
   SSH_MSG_USERAUTH_FAILURE is returned with partial success set to
   false.

   If the user name is not empty, and the requested user does not
   exist, the server MAY disconnect, or MAY send a bogus list of
   acceptable authentications but never accept any.  This makes it
   possible for the server to avoid disclosing information about which
   accounts exist.  In any case, if the user does not exist, the
   authentication request MUST NOT be accepted.

   Any implementation supporting at least one key exchange method which
   conforms to section 1 of this document SHOULD also support the
   "external-keyx" user authentication method, in order to allow user
   authentication to be performed at the same time as key exchange,
   thereby reducing the number of round trips needed for connection
   setup.

5. Null Host Key Algorithm

   The "null" host key algorithm has no associated host key material,
   and provides neither signature nor encryption algorithms.  Thus, it
   can be used only with key exchange methods that do not require any
   public-key operations and do not require the use of host public key
   material.  The key exchange methods described in section 1 of this
   document are examples of such methods.

   This algorithm is used when, as a matter of configuration, the host
   does not have or does not wish to use a public key.  For example, it
   can be used when the administrator has decided as a matter of policy
   to require that all key exchanges be authenticated using Kerberos
   [3], and thus the only permitted key exchange method is the
   GSSAPI-authenticated Diffie-Hellman exchange described above, with
   Kerberos V5 as the underlying GSSAPI mechanism.  In such a
   configuration, the server implementation supports the "ssh-dss" key
   algorithm (as required by [11]), but could be prohibited by
   configuration from using it.  In this situation, the server needs
   some key exchange algorithm to advertise; the "null" algorithm fills
   this purpose.

   Note that the use of the "null" algorithm in this way means that the
   server will not be able to interoperate with clients which do not
   support this algorithm.  This is not a significant problem, since in
   the configuration described, it will also be unable to interoperate
   with implementations that do not support the GSSAPI-authenticated
   key exchange and Kerberos.

   Any implementation supporting at least one key exchange method which
   conforms to section 1 of this document MUST also support the "null"
   host key algorithm.  Servers MUST NOT advertise the "null" host key
   algorithm unless it is the only algorithm advertised.

6. Summary of Message Numbers

   The following message numbers have been defined for use with
   GSSAPI-based key exchange methods:

          #define SSH_MSG_KEXGSS_INIT                       30
          #define SSH_MSG_KEXGSS_CONTINUE                   31
          #define SSH_MSG_KEXGSS_COMPLETE                   32
          #define SSH_MSG_KEXGSS_HOSTKEY                    33
          #define SSH_MSG_KEXGSS_ERROR                      34

   The numbers 30-49 are specific to key exchange and may be redefined
   by other kex methods.

   The following message numbers have been defined for use with the
   'gssapi' user authentication method:

          #define SSH_MSG_USERAUTH_GSSAPI_RESPONSE          60
          #define SSH_MSG_USERAUTH_GSSAPI_TOKEN             61
          #define SSH_MSG_USERAUTH_GSSAPI_EXCHANGE_COMPLETE 63
          #define SSH_MSG_USERAUTH_GSSAPI_ERROR             64

   The numbers 60-79 are specific to user authentication and may be
   redefined by other user auth methods.  Note that in the method
   described in this document, message number 62 is unused.

7. GSSAPI Considerations

7.1 Naming Conventions

   In order to establish a GSSAPI security context, the SSH client
   needs to determine the appropriate targ_name to use in identifying
   the server when calling GSS_Init_sec_context.  For this purpose, the
   GSSAPI mechanism-independent name form for host-based services is
   used, as described in section 4.1 of [2].

   In particular, the targ_name to pass to GSS_Init_sec_context is
   obtained by calling GSS_Import_name with an input_name_type of
   GSS_C_NT_HOSTBASED_SERVICE, and an input_name_string consisting of
   the string "host@" concatenated with the hostname of the SSH server.

7.2 Channel Bindings

   This document recommends that channel bindings SHOULD NOT be
   specified in the calls during context establishment.  This document
   does not specify any standard data to be used as channel bindings
   and the use of network addresses as channel bindings may break SSH
   in environments where it is most useful.

7.3 SPNEGO

   The use of the Simple and Protected GSS-API Negotiation Mechanism
   [8] in conjunction with the authentication and key exchange methods
   described in this document is both unnecessary and undesirable.  As
   a result, mechanisms conforming to this document MUST NOT use SPNEGO
   as the underlying GSSAPI mechanism.

   Since SSH performs its own negotiation of authentication and key
   exchange methods, the negotiation capability of SPNEGO alone does
   not provide any added benefit.  In fact, as described below, it has
   the potential to result in the use of a weaker method than desired.

   Normally, SPNEGO provides the added benefit of protecting the GSSAPI
   mechanism negotiation.  It does this by having the server compute a
   MIC of the list of mechanisms proposed by the client, and then
   checking that value at the client.  In the case of key exchange,
   this protection is not needed because the key exchange methods
   described here already perform an equivalent operation; namely, they
   generate a MIC of the SSH exchange hash, which is a hash of several
   items including the lists of key exchange mechanisms supported by
   both sides.  In the case of user authentication, the protection is
   not needed because the negotiation occurs over a secure channel, and
   the host's identity has already been proved to the user.

   The use of SPNEGO combined with GSSAPI mechanisms used without
   SPNEGO can lead to interoperability problems.  For example, a client
   which supports key exchange using the Kerberos V5 GSSAPI mechanism
   [4] only underneath SPNEGO will not interoperate with a server which
   supports key exchange only using the Kerberos V5 GSSAPI mechanism
   directly.  As a result, allowing GSSAPI mechanisms to be used both
   with and without SPNEGO is undesirable.

   If a client's policy is to first prefer GSSAPI-based key exchange
   method X, then non-GSSAPI method Y, then GSSAPI-based method Z, and
   if a server supports mechanisms Y and Z but not X, then an attempt
   to use SPNEGO to negotiate a GSSAPI mechanism might result in the
   use of method Z when method Y would have been preferable.  As a
   result, the use of SPNEGO could result in the subversion of the
   negotiation algorithm for key exchange methods as described in
   section 5.1 of [11] and/or the negotiation algorithm for user
   authentication methods as described in [12].

8. Security Considerations

   This document describes authentication and key-exchange protocols.
   As such, security considerations are discussed throughout.

   This protocol depends on the SSH protocol itself, the GSSAPI, any
   underlying GSSAPI mechanisms which are used, and any protocols on
   which such mechanisms might depend.  Each of these components plays
   a part in the security of the resulting connection, and each will
   have its own security considerations.

   The key exchange method described in section 1 of this document
   depends on the underlying GSSAPI mechanism to provide both mutual
   authentication and per-message integrity services.  If either of
   these features is not supported by a particular GSSAPI mechanism, or
   by a particular implementation of a GSSAPI mechanism, then the key
   exchange is not secure and MUST fail.

   In order for the "external-keyx" user authentication method to be
   used, it MUST have access to user authentication information
   obtained as a side-effect of the key exchange.  If this information
   is unavailable, the authentication MUST fail.

   Revealing information about the reason for an authentication failure
   may be considered by some sites to be an unacceptable security risk
   for a production environment.  However, having that information
   available can be invaluable for debugging purposes.  Thus, it is
   RECOMMENDED that implementations provide a means for controlling, as
   a matter of policy, whether the SSH_MSG_KEXGSS_ERROR and/or
   SSH_MGS_USERAUTH_GGSAPI_ERROR messages are sent.

9. Acknowledgements

   The authors would like to thank Sam Hartman and Simon Wilkinson for
   their invaluable assistance with this document.

10. Changes the last version

   This section lists important changes since the previous version of
   this internet-draft.  This section should be removed at the time of
   publication of this document as an RFC.

   o  Added  Clarified the SSH_MSG_KEXGSS_ERROR message to allow reporting encoding of
      GSSAPI errors during the key exchange process. host keys in SSH_MSG_KEXGSS_HOSTKEY.

   o  Added  Fixed a wording error in the SSH_MSG_USERAUTH_GSSAPI_ERROR message to allow
      reporting description of GSSAPI errors during the user authentication process.

   o  Clarified exchange hash;
      the handling use of
      GSS_MSG_USERAUTH_GSSAPI_EXCHANGE_COMPLETE when the client has a
      final GSSAPI token to send.

   o  Added references to RFC2279 (UTF-8) and RFC1176 (language tags).

   o  Added a missing paragraph specifying that a server accepting a
      GSSAPI context for empty string as the host key exchange must verify that message
      integrity protection is available in that context. dependent on the
      SSH_MSG_KEXGSS_HOSTKEY message, which is sent by the server and
      received by the client, not the other way around.

References

   [1]   ISO/IEC, "Specification of Abstract Syntax Notation One
         (ASN.1)", ISO/IEC 8824, November 1998.

   [2]   Linn, J., "Generic Security Service Application Program
         Interface Version 2, Update 1", RFC 2743, January 2000.

   [3]   Kohl, J. and C. Neuman, "The Kerberos Network Authentication
         Service (V5)", RFC 1510, September 1993.

   [4]   Linn, J., "The Kerberos Version 5 GSS-API Mechanism", RFC
         1964, June 1996.

   [5]   Rivest, R., "The MD5 Message-Digest Algorithm", RFC 1321,
         April 1992.

   [6]   Freed, N. and N. Borenstein, "Multipurpose Internet Mail
         Extensions (MIME) Part One: Format of Internet Message
         Bodies", RFC 2045, November 1996.

   [7]   Bradner, S., "Key words for use in RFCs to Indicate
         Requirement Levels", RFC 2119, BCP 14, March 1997.

   [8]   Baize, E. and D. Pinkas, "The Simple and Protected GSS-API
         Negotiation Mechanism", RFC 2478, December 1998.

   [9]   Ylonen, T., Kivinen, T., Saarinen, M., Rinne, T. and S.
         Lehtinen, "SSH Protocol Architecture",
         draft-ietf-secsh-architecture-11.txt (work in progress),
         November 2001.

   [10]  Ylonen, T., Kivinen, T., Saarinen, M., Rinne, T. and S.
         Lehtinen, "SSH Connection Protocol",
         draft-ietf-secsh-connect-14.txt (work in progress), November
         2001.

   [11]  Ylonen, T., Kivinen, T., Saarinen, M., Rinne, T. and S.
         Lehtinen, "SSH Transport Layer Protocol",
         draft-ietf-secsh-transport-11.txt (work in progress), November
         2001.

   [12]  Ylonen, T., Kivinen, T., Saarinen, M., Rinne, T. and S.
         Lehtinen, "SSH Authentication Protocol",
         draft-ietf-secsh-userauth-13.txt (work in progress), November
         2001.

   [13]  Yergeau, , "UTF-8, a transformation format of ISO 10646", RFC
         2279, January 1998.

   [14]  Alvestrand, H., "Tags for the Identification of Languages",
         RFC 1766, March 1995.

Authors' Addresses

   Jeffrey Hutzelman
   Carnegie Mellon University
   5000 Forbes Ave
   Pittsburgh, PA  15213
   US

   Phone: +1 412 268 7225
   EMail: jhutz+@cmu.edu
   URI:   http://www.cs.cmu.edu/~jhutz/

   Joseph Salowey
   Cisco Systems
   Bldg 20
   725 Alder Drive
   Milpitas, CA  95035
   US

   Phone: +1 408 525 6381
   EMail: jsalowey@cisco.com

   Joseph Galbraith
   Van Dyke Technologies, Inc.
   4848 Tramway Ridge Dr. NE
   Suite 101
   Albuquerque, NM  87111
   US

   EMail: galb@vandyke.com

   Vol

   Von Welch
   University of Chicago & Argonne National Laboratory
   Distributed Systems Laboratory
   701 E. Washington
   Urbana, IL  61801
   US

   EMail: welch@mcs.anl.gov

Full Copyright Statement

   Copyright (C) The Internet Society (2002). All Rights Reserved.

   This document and translations of it may be copied and furnished to
   others, and derivative works that comment on or otherwise explain it
   or assist in its implementation may be prepared, copied, published
   and distributed, in whole or in part, without restriction of any
   kind, provided that the above copyright notice and this paragraph
   are included on all such copies and derivative works. However, this
   document itself may not be modified in any way, such as by removing
   the copyright notice or references to the Internet Society or other
   Internet organizations, except as needed for the purpose of
   developing Internet standards in which case the procedures for
   copyrights defined in the Internet Standards process must be
   followed, or as required to translate it into languages other than
   English.

   The limited permissions granted above are perpetual and will not be
   revoked by the Internet Society or its successors or assigns.

   This document and the information contained herein is provided on an
   "AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
   TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
   BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
   HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
   MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

Acknowledgement

   Funding for the RFC editor function is currently provided by the
   Internet Society.