Now pac from christian passes since we make hmac checksums always use the raw key

[heimdal.git] / doc / standardisation / draft-ietf-krb-wg-preauth-framework-08.txt

\r
\r
Kerberos Working Group                                            L. Zhu\r
Internet-Draft                                     Microsoft Corporation\r
Updates: 4120 (if approved)                                   S. Hartman\r
Intended status: Standards Track                       Painless Security\r
Expires: January 15, 2009                                  July 14, 2008\r
\r
\r
        A Generalized Framework for Kerberos Pre-Authentication\r
                 draft-ietf-krb-wg-preauth-framework-08\r
\r
Status of this Memo\r
\r
   By submitting this Internet-Draft, each author represents that any\r
   applicable patent or other IPR claims of which he or she is aware\r
   have been or will be disclosed, and any of which he or she becomes\r
   aware will be disclosed, in accordance with Section 6 of BCP 79.\r
\r
   Internet-Drafts are working documents of the Internet Engineering\r
   Task Force (IETF), its areas, and its working groups.  Note that\r
   other groups may also distribute working documents as Internet-\r
   Drafts.\r
\r
   Internet-Drafts are draft documents valid for a maximum of six months\r
   and may be updated, replaced, or obsoleted by other documents at any\r
   time.  It is inappropriate to use Internet-Drafts as reference\r
   material or to cite them other than as "work in progress."\r
\r
   The list of current Internet-Drafts can be accessed at\r
   http://www.ietf.org/ietf/1id-abstracts.txt.\r
\r
   The list of Internet-Draft Shadow Directories can be accessed at\r
   http://www.ietf.org/shadow.html.\r
\r
   This Internet-Draft will expire on January 15, 2009.\r
\r
Abstract\r
\r
   Kerberos is a protocol for verifying the identity of principals\r
   (e.g., a workstation user or a network server) on an open network.\r
   The Kerberos protocol provides a mechanism called pre-authentication\r
   for proving the identity of a principal and for better protecting the\r
   long-term secret of the principal.\r
\r
   This document describes a model for Kerberos pre-authentication\r
   mechanisms.  The model describes what state in the Kerberos request a\r
   pre-authentication mechanism is likely to change.  It also describes\r
   how multiple pre-authentication mechanisms used in the same request\r
   will interact.\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 1]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   This document also provides common tools needed by multiple pre-\r
   authentication mechanisms.  One of these tools is a secure channel\r
   between the client and the KDC with a reply key delivery mechanism;\r
   this secure channel can be used to protect the authentication\r
   exchange thus eliminate offline dictionary attacks.  With these\r
   tools, it is relatively straightforward to chain multiple\r
   authentication mechanisms, utilize a different key management system,\r
   or support a new key agreement algorithm.\r
\r
\r
Table of Contents\r
\r
   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  4\r
   2.  Conventions and Terminology Used in This Document  . . . . . .  5\r
   3.  Model for Pre-Authentication . . . . . . . . . . . . . . . . .  5\r
     3.1.  Information Managed by the Pre-authentication Model  . . .  6\r
     3.2.  Initial Pre-authentication Required Error  . . . . . . . .  8\r
     3.3.  Client to KDC  . . . . . . . . . . . . . . . . . . . . . .  9\r
     3.4.  KDC to Client  . . . . . . . . . . . . . . . . . . . . . . 10\r
   4.  Pre-Authentication Facilities  . . . . . . . . . . . . . . . . 10\r
     4.1.  Client-authentication Facility . . . . . . . . . . . . . . 12\r
     4.2.  Strengthening-reply-key Facility . . . . . . . . . . . . . 12\r
     4.3.  Replacing-reply-key Facility . . . . . . . . . . . . . . . 13\r
     4.4.  KDC-authentication Facility  . . . . . . . . . . . . . . . 14\r
   5.  Requirements for Pre-Authentication Mechanisms . . . . . . . . 14\r
   6.  Tools for Use in Pre-Authentication Mechanisms . . . . . . . . 15\r
     6.1.  Combining Keys . . . . . . . . . . . . . . . . . . . . . . 15\r
     6.2.  Protecting Requests/Responses  . . . . . . . . . . . . . . 16\r
     6.3.  Managing States for the KDC  . . . . . . . . . . . . . . . 17\r
     6.4.  Pre-authentication Set . . . . . . . . . . . . . . . . . . 19\r
     6.5.  Definition of Kerberos FAST Padata . . . . . . . . . . . . 22\r
       6.5.1.  FAST Armors  . . . . . . . . . . . . . . . . . . . . . 23\r
       6.5.2.  FAST Request . . . . . . . . . . . . . . . . . . . . . 24\r
       6.5.3.  FAST Response  . . . . . . . . . . . . . . . . . . . . 28\r
       6.5.4.  Authenticated Kerberos Error Messages using\r
               Kerberos FAST  . . . . . . . . . . . . . . . . . . . . 30\r
       6.5.5.  The Encrypted Challenge FAST Factor  . . . . . . . . . 31\r
     6.6.  Authentication Strength Indication . . . . . . . . . . . . 32\r
   7.  IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 33\r
   8.  Security Considerations  . . . . . . . . . . . . . . . . . . . 33\r
   9.  Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 34\r
   10. References . . . . . . . . . . . . . . . . . . . . . . . . . . 34\r
     10.1. Normative References . . . . . . . . . . . . . . . . . . . 34\r
     10.2. Informative References . . . . . . . . . . . . . . . . . . 34\r
   Appendix A.  Change History  . . . . . . . . . . . . . . . . . . . 35\r
     A.1.  Changes since 07 . . . . . . . . . . . . . . . . . . . . . 35\r
     A.2.  Changes since 06 . . . . . . . . . . . . . . . . . . . . . 35\r
   Appendix B.  ASN.1 module  . . . . . . . . . . . . . . . . . . . . 35\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 2]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 39\r
   Intellectual Property and Copyright Statements . . . . . . . . . . 40\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 3]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
1.  Introduction\r
\r
   The core Kerberos specification [RFC4120] treats pre-authentication\r
   data as an opaque typed hole in the messages to the KDC that may\r
   influence the reply key used to encrypt the KDC reply.  This\r
   generality has been useful: pre-authentication data is used for a\r
   variety of extensions to the protocol, many outside the expectations\r
   of the initial designers.  However, this generality makes designing\r
   more common types of pre-authentication mechanisms difficult.  Each\r
   mechanism needs to specify how it interacts with other mechanisms.\r
   Also, problems like combining a key with the long-term secret or\r
   proving the identity of the user are common to multiple mechanisms.\r
   Where there are generally well-accepted solutions to these problems,\r
   it is desirable to standardize one of these solutions so mechanisms\r
   can avoid duplication of work.  In other cases, a modular approach to\r
   these problems is appropriate.  The modular approach will allow new\r
   and better solutions to common pre-authentication problems to be used\r
   by existing mechanisms as they are developed.\r
\r
   This document specifies a framework for Kerberos pre-authentication\r
   mechanisms.  It defines the common set of functions that pre-\r
   authentication mechanisms perform as well as how these functions\r
   affect the state of the request and reply.  In addition several\r
   common tools needed by pre-authentication mechanisms are provided.\r
   Unlike [RFC3961], this framework is not complete--it does not\r
   describe all the inputs and outputs for the pre-authentication\r
   mechanisms.  Pre-Authentication mechanism designers should try to be\r
   consistent with this framework because doing so will make their\r
   mechanisms easier to implement.  Kerberos implementations are likely\r
   to have plugin architectures for pre-authentication; such\r
   architectures are likely to support mechanisms that follow this\r
   framework plus commonly used extensions.\r
\r
   One of these common tools is the flexible authentication secure\r
   tunneling (FAST) padata type.  FAST provides a protected channel\r
   between the client and the KDC, and it can optionally deliver a reply\r
   key within the protected channel.  Based on FAST, pre-authentication\r
   mechanisms can extend Kerberos with ease, to support, for example,\r
   password authenticated key exchange (PAKE) protocols with zero\r
   knowledge password proof (ZKPP) [EKE] [IEEE1363.2].  Any pre-\r
   authentication mechanism can be encapsulated in the FAST messages as\r
   defined in Section 6.5.  A pre-authentication type carried within\r
   FAST is called a FAST factor.  Creating a FAST factor is the easiest\r
   path to create a new pre-authentication mechanism.  FAST factors are\r
   significantly easier to analyze from a security standpoint than other\r
   pre-authentication mechanisms.\r
\r
   Mechanism designers should design FAST factors, instead of new pre-\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 4]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   authentication mechanisms outside of FAST.\r
\r
\r
2.  Conventions and Terminology Used in This Document\r
\r
   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",\r
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this\r
   document are to be interpreted as described in [RFC2119].\r
\r
   The word padata is used as a shorthand for pre-authentication data.\r
\r
   A conversation is the set of all authentication messages exchanged\r
   between the client and the KDCs in order to authenticate the client\r
   principal.  A conversation as defined here consists of all messages\r
   that are necessary to complete the authentication between the client\r
   and the KDC.\r
\r
   Lastly, this document should be read only after reading the documents\r
   describing the Kerberos cryptography framework [RFC3961] and the core\r
   Kerberos protocol [RFC4120].  This document may freely use\r
   terminology and notation from these documents without reference or\r
   further explanation.\r
\r
\r
3.  Model for Pre-Authentication\r
\r
   When a Kerberos client wishes to obtain a ticket using the\r
   authentication server, it sends an initial Authentication Service\r
   (AS) request.  If pre-authentication is required but not being used,\r
   then the KDC will respond with a KDC_ERR_PREAUTH_REQUIRED error.\r
   Alternatively, if the client knows what pre-authentication to use, it\r
   MAY optimize away a round-trip and send an initial request with\r
   padata included in the initial request.  If the client includes the\r
   padata computed using the wrong pre-authentication mechanism or\r
   incorrect keys, the KDC MAY return KDC_ERR_PREAUTH_FAILED with no\r
   indication of what padata should have been included.  In that case,\r
   the client MUST retry with no padata and examine the error data of\r
   the KDC_ERR_PREAUTH_REQUIRED error.  If the KDC includes pre-\r
   authentication information in the accompanying error data of\r
   KDC_ERR_PREAUTH_FAILED, the client SHOULD process the error data, and\r
   then retry.\r
\r
   The conventional KDC maintains no state between two requests;\r
   subsequent requests may even be processed by a different KDC.  On the\r
   other hand, the client treats a series of exchanges with KDCs as a\r
   single conversation.  Each exchange accumulates state and hopefully\r
   brings the client closer to a successful authentication.\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 5]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   These models for state management are in apparent conflict.  For many\r
   of the simpler pre-authentication scenarios, the client uses one\r
   round trip to find out what mechanisms the KDC supports.  Then the\r
   next request contains sufficient pre-authentication for the KDC to be\r
   able to return a successful reply.  For these simple scenarios, the\r
   client only sends one request with pre-authentication data and so the\r
   conversation is trivial.  For more complex conversations, the KDC\r
   needs to provide the client with a cookie to include in future\r
   requests to capture the current state of the authentication session.\r
   Handling of multiple round-trip mechanisms is discussed in\r
   Section 6.3.\r
\r
   This framework specifies the behavior of Kerberos pre-authentication\r
   mechanisms used to identify users or to modify the reply key used to\r
   encrypt the KDC reply.  The PA-DATA typed hole may be used to carry\r
   extensions to Kerberos that have nothing to do with proving the\r
   identity of the user or establishing a reply key.  Such extensions\r
   are outside the scope of this framework.  However mechanisms that do\r
   accomplish these goals should follow this framework.\r
\r
   This framework specifies the minimum state that a Kerberos\r
   implementation needs to maintain while handling a request in order to\r
   process pre-authentication.  It also specifies how Kerberos\r
   implementations process the padata at each step of the AS request\r
   process.\r
\r
3.1.  Information Managed by the Pre-authentication Model\r
\r
   The following information is maintained by the client and KDC as each\r
   request is being processed:\r
\r
   o  The reply key used to encrypt the KDC reply\r
\r
   o  How strongly the identity of the client has been authenticated\r
\r
   o  Whether the reply key has been used in this conversation\r
\r
   o  Whether the reply key has been replaced in this conversation\r
\r
   o  Whether the contents of the KDC reply can be verified by the\r
      client principal\r
\r
\r
   Conceptually, the reply key is initially the long-term key of the\r
   principal.  However, principals can have multiple long-term keys\r
   because of support for multiple encryption types, salts and\r
   string2key parameters.  As described in Section 5.2.7.5 of the\r
   Kerberos protocol [RFC4120], the KDC sends PA-ETYPE-INFO2 to notify\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 6]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   the client what types of keys are available.  Thus in full\r
   generality, the reply key in the pre-authentication model is actually\r
   a set of keys.  At the beginning of a request, it is initialized to\r
   the set of long-term keys advertised in the PA-ETYPE-INFO2 element on\r
   the KDC.  If multiple reply keys are available, the client chooses\r
   which one to use.  Thus the client does not need to treat the reply\r
   key as a set.  At the beginning of a request, the client picks a\r
   reply key to use.\r
\r
   KDC implementations MAY choose to offer only one key in the PA-ETYPE-\r
   INFO2 element.  Since the KDC already knows the client's list of\r
   supported enctypes from the request, no interoperability problems are\r
   created by choosing a single possible reply key.  This way, the KDC\r
   implementation avoids the complexity of treating the reply key as a\r
   set.\r
\r
   When the padata in the request is verified by the KDC, then the\r
   client is known to have that key, therefore the KDC SHOULD pick the\r
   same key as the reply key.\r
\r
   At the beginning of handling a message on both the client and the\r
   KDC, the client's identity is not authenticated.  A mechanism may\r
   indicate that it has successfully authenticated the client's\r
   identity.  This information is useful to keep track of on the client\r
   in order to know what pre-authentication mechanisms should be used.\r
   The KDC needs to keep track of whether the client is authenticated\r
   because the primary purpose of pre-authentication is to authenticate\r
   the client identity before issuing a ticket.  The handling of\r
   authentication strength using various authentication mechanisms is\r
   discussed in Section 6.6.\r
\r
   Initially the reply key has not been used.  A pre-authentication\r
   mechanism that uses the reply key to encrypt or checksum some data in\r
   the generation of new keys MUST indicate that the reply key is used.\r
   This state is maintained by the client and the KDC to enforce the\r
   security requirement stated in Section 4.3 that the reply key cannot\r
   be replaced after it is used.\r
\r
   Initially the reply key has not been replaced.  If a mechanism\r
   implements the Replace Reply Key facility discussed in Section 4.3,\r
   then the state MUST be updated to indicate that the reply key has\r
   been replaced.  Once the reply key has been replaced, knowledge of\r
   the reply key is insufficient to authenticate the client.  The reply\r
   key is marked replaced in exactly the same situations as the KDC\r
   reply is marked as not being verified to the client principal.\r
   However, while mechanisms can verify the KDC reply to the client,\r
   once the reply key is replaced, then the reply key remains replaced\r
   for the remainder of the conversation.\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 7]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   Without pre-authentication, the client knows that the KDC reply is\r
   authentic and has not been modified because it is encrypted in a\r
   long-term key of the client.  Only the KDC and the client know that\r
   key.  So at the start of a conversation, the KDC reply is presumed to\r
   be verified using the client principal's long-term key.  Any pre-\r
   authentication mechanism that sets a new reply key not based on the\r
   principal's long-term secret MUST either verify the KDC reply some\r
   other way or indicate that the reply is not verified.  If a mechanism\r
   indicates that the reply is not verified then the client\r
   implementation MUST return an error unless a subsequent mechanism\r
   verifies the reply.  The KDC needs to track this state so it can\r
   avoid generating a reply that is not verified.\r
\r
   The typical Kerberos request does not provide a way for the client\r
   machine to know that it is talking to the correct KDC.  Someone who\r
   can inject packets into the network between the client machine and\r
   the KDC and who knows the password that the user will give to the\r
   client machine can generate a KDC reply that will decrypt properly.\r
   So, if the client machine needs to authenticate that the user is in\r
   fact the named principal, then the client machine needs to do a TGS\r
   request for itself as a service.  Some pre-authentication mechanisms\r
   may provide a way for the client to authenticate the KDC.  Examples\r
   of this include signing the reply that can be verified using a well-\r
   known public key or providing a ticket for the client machine as a\r
   service.\r
\r
3.2.  Initial Pre-authentication Required Error\r
\r
   Typically a client starts a conversation by sending an initial\r
   request with no pre-authentication.  If the KDC requires pre-\r
   authentication, then it returns a KDC_ERR_PREAUTH_REQUIRED message.\r
   After the first reply with the KDC_ERR_PREAUTH_REQUIRED error code,\r
   the KDC returns the error code KDC_ERR_MORE_PREAUTH_DATA_NEEDED\r
   (defined in Section 6.3) for pre-authentication configurations that\r
   use multi-round-trip mechanisms; see Section 3.4 for details of that\r
   case.\r
\r
   The KDC needs to choose which mechanisms to offer the client.  The\r
   client needs to be able to choose what mechanisms to use from the\r
   first message.  For example consider the KDC that will accept\r
   mechanism A followed by mechanism B or alternatively the single\r
   mechanism C. A client that supports A and C needs to know that it\r
   should not bother trying A.\r
\r
   Mechanisms can either be sufficient on their own or can be part of an\r
   authentication set--a group of mechanisms that all need to\r
   successfully complete in order to authenticate a client.  Some\r
   mechanisms may only be useful in authentication sets; others may be\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 8]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   useful alone or in authentication sets.  For the second group of\r
   mechanisms, KDC policy dictates whether the mechanism will be part of\r
   an authentication set or offered alone.  For each mechanism that is\r
   offered alone, the KDC includes the pre-authentication type ID of the\r
   mechanism in the padata sequence returned in the\r
   KDC_ERR_PREAUTH_REQUIRED error.\r
\r
   The KDC SHOULD NOT send data that is encrypted in the long-term\r
   password-based key of the principal.  Doing so has the same security\r
   exposures as the Kerberos protocol without pre-authentication.  There\r
   are few situations where pre-authentication is desirable and where\r
   the KDC needs to expose cipher text encrypted in a weak key before\r
   the client has proven knowledge of that key.\r
\r
3.3.  Client to KDC\r
\r
   This description assumes that a client has already received a\r
   KDC_ERR_PREAUTH_REQUIRED from the KDC.  If the client performs\r
   optimistic pre-authentication then the client needs to optimistically\r
   guess values for the information it would normally receive from that\r
   error response.\r
\r
   The client starts by initializing the pre-authentication state as\r
   specified.  It then processes the padata in the\r
   KDC_ERR_PREAUTH_REQUIRED.\r
\r
   When processing the response to the KDC_ERR_PREAUTH_REQUIRED, the\r
   client MAY ignore any padata it chooses unless doing so violates a\r
   specification to which the client conforms.  Clients conforming to\r
   this specification MUST NOT ignore the padata defined in Section 6.3.\r
   Clients SHOULD process padata unrelated to this framework or other\r
   means of authenticating the user.  Clients SHOULD choose one\r
   authentication set or mechanism that could lead to authenticating the\r
   user and ignore the rest.  Since the list of mechanisms offered by\r
   the KDC is in the decreasing preference order, clients typically\r
   choose the first mechanism or authentication set that the client can\r
   usefully perform.  If a client chooses to ignore a padata it MUST NOT\r
   process the padata, allow the padata to affect the pre-authentication\r
   state, nor respond to the padata.\r
\r
   For each padata the client chooses to process, the client processes\r
   the padata and modifies the pre-authentication state as required by\r
   that mechanism.  Padata are processed in the order received from the\r
   KDC.\r
\r
   After processing the padata in the KDC error, the client generates a\r
   new request.  It processes the pre-authentication mechanisms in the\r
   order in which they will appear in the next request, updating the\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009                [Page 9]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   state as appropriate.  The request is sent when it is complete.\r
\r
3.4.  KDC to Client\r
\r
   When a KDC receives an AS request from a client, it needs to\r
   determine whether it will respond with an error or an AS reply.\r
   There are many causes for an error to be generated that have nothing\r
   to do with pre-authentication; they are discussed in the core\r
   Kerberos specification.\r
\r
   From the standpoint of evaluating the pre-authentication, the KDC\r
   first starts by initializing the pre-authentication state.  It then\r
   processes the padata in the request.  As mentioned in Section 3.3,\r
   the KDC MAY ignore padata that is inappropriate for the configuration\r
   and MUST ignore padata of an unknown type.  The KDC MUST NOT ignore\r
   padata of types used in previous messages.  For example, if a KDC\r
   issues a KDC_ERR_PREAUTH_REQUIRED error including padata of type x,\r
   then the KDC cannot ignore padata of type x received in an AS-REQ\r
   message from the client.\r
\r
   At this point the KDC decides whether it will issue an error or a\r
   reply.  Typically a KDC will issue a reply if the client's identity\r
   has been authenticated to a sufficient degree.\r
\r
   In the case of a KDC_ERR_MORE_PREAUTH_DATA_NEEDED error, the KDC\r
   first starts by initializing the pre-authentication state.  Then it\r
   processes any padata in the client's request in the order provided by\r
   the client.  Mechanisms that are not understood by the KDC are\r
   ignored.  Next, it generates padata for the error response, modifying\r
   the pre-authentication state appropriately as each mechanism is\r
   processed.  The KDC chooses the order in which it will generate\r
   padata (and thus the order of padata in the response), but it needs\r
   to modify the pre-authentication state consistently with the choice\r
   of order.  For example, if some mechanism establishes an\r
   authenticated client identity, then the subsequent mechanisms in the\r
   generated response receive this state as input.  After the padata is\r
   generated, the error response is sent.  Typically the errors with the\r
   code KDC_ERR_MORE_PREAUTH_DATA_NEEDED in a converstation will include\r
   KDC state as discussed in Section 6.3.\r
\r
   To generate a final reply, the KDC generates the padata modifying the\r
   pre-authentication state as necessary.  Then it generates the final\r
   response, encrypting it in the current pre-authentication reply key.\r
\r
\r
4.  Pre-Authentication Facilities\r
\r
   Pre-Authentication mechanisms can be thought of as providing various\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 10]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   conceptual facilities.  This serves two useful purposes.  First,\r
   mechanism authors can choose only to solve one specific small\r
   problem.  It is often useful for a mechanism designed to offer key\r
   management not to directly provide client authentication but instead\r
   to allow one or more other mechanisms to handle this need.  Secondly,\r
   thinking about the abstract services that a mechanism provides yields\r
   a minimum set of security requirements that all mechanisms providing\r
   that facility must meet.  These security requirements are not\r
   complete; mechanisms will have additional security requirements based\r
   on the specific protocol they employ.\r
\r
   A mechanism is not constrained to only offering one of these\r
   facilities.  While such mechanisms can be designed and are sometimes\r
   useful, many pre-authentication mechanisms implement several\r
   facilities.  By combining multiple facilities in a single mechanism,\r
   it is often easier to construct a secure, simple solution than by\r
   solving the problem in full generality.  Even when mechanisms provide\r
   multiple facilities, they need to meet the security requirements for\r
   all the facilities they provide.  If the FAST factor approach is\r
   used, it is likely that one or a small number of facilities can be\r
   provided by a single mechanism without complicating the security\r
   analysis.\r
\r
   According to Kerberos extensibility rules (Section 1.5 of the\r
   Kerberos specification [RFC4120]), an extension MUST NOT change the\r
   semantics of a message unless a recipient is known to understand that\r
   extension.  Because a client does not know that the KDC supports a\r
   particular pre-authentication mechanism when it sends an initial\r
   request, a pre-authentication mechanism MUST NOT change the semantics\r
   of the request in a way that will break a KDC that does not\r
   understand that mechanism.  Similarly, KDCs MUST NOT send messages to\r
   clients that affect the core semantics unless the client has\r
   indicated support for the message.\r
\r
   The only state in this model that would break the interpretation of a\r
   message is changing the expected reply key.  If one mechanism changed\r
   the reply key and a later mechanism used that reply key, then a KDC\r
   that interpreted the second mechanism but not the first would fail to\r
   interpret the request correctly.  In order to avoid this problem,\r
   extensions that change core semantics are typically divided into two\r
   parts.  The first part proposes a change to the core semantic--for\r
   example proposes a new reply key.  The second part acknowledges that\r
   the extension is understood and that the change takes effect.\r
   Section 4.2 discusses how to design mechanisms that modify the reply\r
   key to be split into a proposal and acceptance without requiring\r
   additional round trips to use the new reply key in subsequent pre-\r
   authentication.  Other changes in the state described in Section 3.1\r
   can safely be ignored by a KDC that does not understand a mechanism.\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 11]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   Mechanisms that modify the behavior of the request outside the scope\r
   of this framework need to carefully consider the Kerberos\r
   extensibility rules to avoid similar problems.\r
\r
4.1.  Client-authentication Facility\r
\r
   The client authentication facility proves the identity of a user to\r
   the KDC before a ticket is issued.  Examples of mechanisms\r
   implementing this facility include the encrypted timestamp facility\r
   defined in Section 5.2.7.2 of the Kerberos specification [RFC4120].\r
   Mechanisms that provide this facility are expected to mark the client\r
   as authenticated.\r
\r
   Mechanisms implementing this facility SHOULD require the client to\r
   prove knowledge of the reply key before transmitting a successful KDC\r
   reply.  Otherwise, an attacker can intercept the pre-authentication\r
   exchange and get a reply to attack.  One way of proving the client\r
   knows the reply key is to implement the Replace Reply Key facility\r
   along with this facility.  The PKINIT mechanism [RFC4556] implements\r
   Client Authentication alongside Replace Reply Key.\r
\r
   If the reply key has been replaced, then mechanisms such as\r
   encrypted-timestamp that rely on knowledge of the reply key to\r
   authenticate the client MUST NOT be used.\r
\r
4.2.  Strengthening-reply-key Facility\r
\r
   Particularly, when dealing with keys based on passwords, it is\r
   desirable to increase the strength of the key by adding additional\r
   secrets to it.  Examples of sources of additional secrets include the\r
   results of a Diffie-Hellman key exchange or key bits from the output\r
   of a smart card [KRB-WG.SAM].  Typically these additional secrets can\r
   be first combined with the existing reply key and then converted to a\r
   protocol key using tools defined in Section 6.1.\r
\r
   Typically a mechanism implementing this facility will know that the\r
   other side of the exchange supports the facility before the reply key\r
   is changed.  For example, a mechanism might need to learn the\r
   certificate for a KDC before encrypting a new key in the public key\r
   belonging to that certificate.  However, if a mechanism implementing\r
   this facility wishes to modify the reply key before knowing that the\r
   other party in the exchange supports the mechanism, it proposes\r
   modifying the reply key.  The other party then includes a message\r
   indicating that the proposal is accepted if it is understood and\r
   meets policy.  In many cases it is desirable to use the new reply key\r
   for client authentication and for other facilities.  Waiting for the\r
   other party to accept the proposal and actually modify the reply key\r
   state would add an additional round trip to the exchange.  Instead,\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 12]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   mechanism designers are encouraged to include a typed hole for\r
   additional padata in the message that proposes the reply key change.\r
   The padata included in the typed hole are generated assuming the new\r
   reply key.  If the other party accepts the proposal, then these\r
   padata are considered as an inner level.  As with the outer level,\r
   one authentication set or mechanism is typically chosen for client\r
   authentication, along with auxiliary mechanisms such as KDC cookies,\r
   and other mechanisms are ignored.  When mechanisms include such a\r
   container, the hint provided for use in authentication sets MUST\r
   contain a sequence of inner mechanisms along with hints for those\r
   mechanisms.  The party generating the proposal can determine whether\r
   the padata were processed based on whether the proposal for the reply\r
   key is accepted.\r
\r
   The specific formats of the proposal message, including where padata\r
   are included is a matter for the mechanism specification.  Similarly,\r
   the format of the message accepting the proposal is mechanism-\r
   specific.\r
\r
   Mechanisms implementing this facility and including a typed hole for\r
   additional padata MUST checksum that padata using a keyed checksum or\r
   encrypt the padata.  This requirement protects against modification\r
   of the contents of the typed hole.  By modifying these contents an\r
   attacker might be able to choose which mechanism is used to\r
   authenticate the client, or to convince a party to provide text\r
   encrypted in a key that the attacker had manipulated.  It is\r
   important that mechanisms strengthen the reply key enough that using\r
   it to checksum padata is appropriate.\r
\r
4.3.  Replacing-reply-key Facility\r
\r
   The Replace Reply Key facility replaces the key in which a successful\r
   AS reply will be encrypted.  This facility can only be used in cases\r
   where knowledge of the reply key is not used to authenticate the\r
   client.  The new reply key MUST be communicated to the client and the\r
   KDC in a secure manner.  Mechanisms implementing this facility MUST\r
   mark the reply key as replaced in the pre-authentication state.\r
   Mechanisms implementing this facility MUST either provide a mechanism\r
   to verify the KDC reply to the client or mark the reply as unverified\r
   in the pre-authentication state.  Mechanisms implementing this\r
   facility SHOULD NOT be used if a previous mechanism has used the\r
   reply key.\r
\r
   As with the strengthening-reply-key facility, Kerberos extensibility\r
   rules require that the reply key not be changed unless both sides of\r
   the exchange understand the extension.  In the case of this facility\r
   it will likely be the case for both sides to know that the facility\r
   is available by the time that the new key is available to be used.\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 13]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   However, mechanism designers can use a container for padata in a\r
   proposal message as discussed in Section 4.2 if appropriate.\r
\r
4.4.  KDC-authentication Facility\r
\r
   This facility verifies that the reply comes from the expected KDC.\r
   In traditional Kerberos, the KDC and the client share a key, so if\r
   the KDC reply can be decrypted then the client knows that a trusted\r
   KDC responded.  Note that the client machine cannot trust the client\r
   unless the machine is presented with a service ticket for it\r
   (typically the machine can retrieve this ticket by itself).  However,\r
   if the reply key is replaced, some mechanism is required to verify\r
   the KDC.  Pre-authentication mechanisms providing this facility allow\r
   a client to determine that the expected KDC has responded even after\r
   the reply key is replaced.  They mark the pre-authentication state as\r
   having been verified.\r
\r
\r
5.  Requirements for Pre-Authentication Mechanisms\r
\r
   This section lists requirements for specifications of pre-\r
   authentication mechanisms.\r
\r
   For each message in the pre-authentication mechanism, the\r
   specification describes the pa-type value to be used and the contents\r
   of the message.  The processing of the message by the sender and\r
   recipient is also specified.  This specification needs to include all\r
   modifications to the pre-authentication state.\r
\r
   Generally mechanisms have a message that can be sent in the error\r
   data of the KDC_ERR_PREAUTH_REQUIRED error message or in an\r
   authentication set.  If the client needs information such as trusted\r
   certificate authorities in order to determine if it can use the\r
   mechanism, then this information should be in that message.  In\r
   addition, such mechanisms should also define a pa-hint to be included\r
   in authentication sets.  Often, the same information included in the\r
   padata-value is appropriate to include in the pa-hint (as defined in\r
   Section 6.4).\r
\r
   In order to ease security analysis the mechanism specification should\r
   describe what facilities from this document are offered by the\r
   mechanism.  For each facility, the security consideration section of\r
   the mechanism specification should show that the security\r
   requirements of that facility are met.  This requirement is\r
   applicable to any FAST factor that provides authentication\r
   information.\r
\r
   Significant problems have resulted in the specification of Kerberos\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 14]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   protocols because much of the KDC exchange is not protected against\r
   authentication.  The security considerations section should discuss\r
   unauthenticated plaintext attacks.  It should either show that\r
   plaintext is protected or discuss what harm an attacker could do by\r
   modifying the plaintext.  It is generally acceptable for an attacker\r
   to be able to cause the protocol negotiation to fail by modifying\r
   plaintext.  More significant attacks should be evaluated carefully.\r
\r
   As discussed in Section 6.3, there is no guarantee that a client will\r
   use the same KDCs for all messages in a conversation.  The mechanism\r
   specification needs to show why the mechanism is secure in this\r
   situation.  The hardest problem to deal with, especially for\r
   challenge/response mechanisms is to make sure that the same response\r
   cannot be replayed against two KDCs while allowing the client to talk\r
   to any KDC.\r
\r
\r
6.  Tools for Use in Pre-Authentication Mechanisms\r
\r
   This section describes common tools needed by multiple pre-\r
   authentication mechanisms.  By using these tools mechanism designers\r
   can use a modular approach to specify mechanism details and ease\r
   security analysis.\r
\r
6.1.  Combining Keys\r
\r
   Frequently a weak key needs to be combined with a stronger key before\r
   use.  For example, passwords are typically limited in size and\r
   insufficiently random, therefore it is desirable to increase the\r
   strength of the keys based on passwords by adding additional secrets.\r
   Additional source of secrecy may come from hardware tokens.\r
\r
   This section provides standard ways to combine two keys into one.\r
\r
   KRB-FX-CF1() is defined to combine two pass-phrases.\r
\r
       KRB-FX-CF1(UTF-8 string, UTF-8 string) -> (UTF-8 string)\r
       KRB-FX-CF1(x, y) -> x || y\r
\r
   Where || denotes concatenation.  The strength of the final key is\r
   roughly the total strength of the individual keys being combined\r
   assuming that the string_to_key() function [RFC3961] uses all its\r
   input evenly.\r
\r
   An example usage of KRB-FX-CF1() is when a device provides random but\r
   short passwords, the password is often combined with a personal\r
   identification number (PIN).  The password and the PIN can be\r
   combined using KRB-FX-CF1().\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 15]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   KRB-FX-CF2() combines two protocol keys based on the pseudo-random()\r
   function defined in [RFC3961].\r
\r
   Given two input keys, K1 and K2, where K1 and K2 can be of two\r
   different enctypes, the output key of KRB-FX-CF2(), K3, is derived as\r
   follows:\r
\r
       KRB-FX-CF2(protocol key, protocol key, octet string,\r
                 octet string)  ->  (protocol key)\r
\r
       PRF+(K1, pepper1) -> octet-string-1\r
       PRF+(K2, pepper2) -> octet-string-2\r
       KRB-FX-CF2(K1, K2, pepper1, pepper2) ->\r
              random-to-key(octet-string-1 ^ octet-string-2)\r
\r
   Where ^ denotes the exclusive-OR operation.  PRF+() is defined as\r
   follows:\r
\r
    PRF+(protocol key, octet string) -> (octet string)\r
\r
    PRF+(key, shared-info) -> pseudo-random( key,  1 || shared-info ) ||\r
                  pseudo-random( key, 2 || shared-info ) ||\r
                  pseudo-random( key, 3 || shared-info ) || ...\r
\r
   Here the counter value 1, 2, 3 and so on are encoded as a one-octet\r
   integer.  The pseudo-random() operation is specified by the enctype\r
   of the protocol key.  PRF+() uses the counter to generate enough bits\r
   as needed by the random-to-key() [RFC3961] function for the\r
   encryption type specified for the resulting key; unneeded bits are\r
   removed from the tail.\r
\r
   Mechanism designers MUST specify the values for the input parameter\r
   pepper1 and pepper2 when combining two keys using KRB-FX-CF2().  The\r
   pepper1 and pepper2 MUST be distinct so that if the two keys being\r
   combined are the same, the resulting key is not a trivial key.\r
\r
6.2.  Protecting Requests/Responses\r
\r
   Mechanism designers SHOULD protect clear text portions of pre-\r
   authentication data.  Various denial of service attacks and downgrade\r
   attacks against Kerberos are possible unless plaintexts are somehow\r
   protected against modification.  An early design goal of Kerberos\r
   Version 5 [RFC4120] was to avoid encrypting more of the\r
   authentication exchange that was required.  (Version 4 doubly-\r
   encrypted the encrypted part of a ticket in a KDC reply, for\r
   example.)  This minimization of encryption reduces the load on the\r
   KDC and busy servers.  Also, during the initial design of Version 5,\r
   the existence of legal restrictions on the export of cryptography\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 16]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   made it desirable to minimize of the number of uses of encryption in\r
   the protocol.  Unfortunately, performing this minimization created\r
   numerous instances of unauthenticated security-relevant plaintext\r
   fields.\r
\r
   If there is more than one roundtrip for an authentication exchange,\r
   mechanism designers need to allow either the client or the KDC to\r
   provide a checksum of all the messages exchanged on the wire in the\r
   conversation, and the checksum is then verified by the receiver.\r
\r
   New mechanisms MUST NOT be hard-wired to use a specific algorithm.\r
\r
   Primitives defined in [RFC3961] are RECOMMENDED for integrity\r
   protection and confidentiality.  Mechanisms based on these primitives\r
   are crypto-agile as the result of using [RFC3961] along with\r
   [RFC4120].  The advantage afforded by crypto-agility is the ability\r
   to avoid a multi-year standardization and deployment cycle to fix a\r
   problem that is specific to a particular algorithm, when real attacks\r
   do arise against that algorithm.\r
\r
   Note that data used by FAST factors (defined in Section 6.5) is\r
   encrypted in a protected channel, thus they do not share the un-\r
   authenticated-text issues with mechanisms designed as full-blown pre-\r
   authentication mechanisms.\r
\r
6.3.  Managing States for the KDC\r
\r
   Kerberos KDCs are stateless.  There is no requirement that clients\r
   will choose the same KDC for the second request in a conversation.\r
   Proxies or other intermediate nodes may also influence KDC selection.\r
   So, each request from a client to a KDC must include sufficient\r
   information that the KDC can regenerate any needed state.  This is\r
   accomplished by giving the client a potentially long opaque cookie in\r
   responses to include in future requests in the same conversation.\r
   The KDC MAY respond that a conversation is too old and needs to\r
   restart by responding with a KDC_ERR_PREAUTH_EXPIRED error.\r
\r
       KDC_ERR_PREAUTH_EXPIRED            TBA\r
\r
   When a client receives this error, the client SHOULD abort the\r
   existing conversation, and restart a new one.\r
\r
   An example, where more than one message from the client is needed, is\r
   when the client is authenticated based on a challenge-response\r
   scheme.  In that case, the KDC needs to keep track of the challenge\r
   issued for a client authentication request.\r
\r
   The PA-FX-COOKIE pdata type is defined in this section to facilitate\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 17]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   state management.  This padata is sent by the KDC when the KDC\r
   requires state for a future transaction.  The client includes this\r
   opaque token in the next message in the conversation.  The token may\r
   be relatively large; clients MUST be prepared for tokens somewhat\r
   larger than the size of all messages in a conversation.\r
\r
       PA_FX_COOKIE                       TBA\r
           -- Stateless cookie that is not tied to a specific KDC.\r
\r
   The corresponding padata-value field [RFC4120] contains the\r
   Distinguished Encoding Rules (DER) [X60] [X690] encoding of the\r
   following Abstract Syntax Notation One (ASN.1) type PA-FX-COOKIE:\r
\r
      PA-FX-COOKIE ::= SEQUENCE {\r
          conversationId  [0] OCTET STRING,\r
             -- Contains the identifier of this conversation. This field\r
             -- must contain the same value for all the messages\r
             -- within the same conversation.\r
          enc-binding-key [1] EncryptedData OPTIONAL,\r
                          -- EncryptionKey --\r
             -- This field is present when and only when a FAST\r
             -- padata as defined in Section 6.5 is included.\r
             -- The encrypted data, when decrypted, contains an\r
             -- EncryptionKey structure.\r
             -- This encryption key is encrypted using the armor key\r
             -- (defined in Section 6.5.1), and the key usage for the\r
             -- encryption is KEY_USAGE_FAST_BINDING_KEY.\r
             -- Present only once in a converstation.\r
          cookie          [2] OCTET STRING OPTIONAL,\r
             -- Opaque data, for use to associate all the messages in\r
             -- a single conversation between the client and the KDC.\r
             -- This is generated by the KDC and the client MUST copy\r
             -- the exact cookie encapsulated in a PA_FX_COOKIE data\r
             -- element into the next message of the same conversation.\r
          ...\r
      }\r
      KEY_USAGE_FAST_BINDING_KEY         TBA\r
\r
   The conversationId field contains a sufficiently-long rand number\r
   that uniquely identifies the conversation.  If a PA_FX_COOKIE padata\r
   is present in one message, a PA_FX_COOKIE structure MUST be present\r
   in all subsequent messages of the same converstation between the\r
   client and the KDC, with the same conversationId value.\r
\r
   The enc-binding-key field is present when and only when a FAST padata\r
   (defined in Section 6.5) is included.  The enc-binding-key field is\r
   present only once in a conversation.  It MUST be ignored if it is\r
   present in a subsequent message of the same conversation.  The\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 18]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   encrypted data, when decrypted, contains an EncryptionKey structure\r
   that is called the binding key.  The binding key is encrypted using\r
   the armor key (defined in Section 6.5.1), and the key usage for the\r
   encryption is KEY_USAGE_FAST_BINDING_KEY.\r
\r
   If a Kerberos FAST padata as defined in Section 6.5 is included in\r
   one message, it MUST be included in all subsequent messages of the\r
   same conversation.\r
\r
   When FAST padata as defined Section 6.5 is included, the PA-FX-COOKIE\r
   padata MUST be included.\r
\r
   The cookie token is generated by the KDC and the client MUST copy the\r
   exact cookie encapsulated in a PA_FX_COOKIE data element into the\r
   next message of the same conversation.  The content of the cookie\r
   field is a local matter of the KDC.  However the KDC MUST construct\r
   the cookie token in such a manner that a malicious client cannot\r
   subvert the authentication process by manipulating the token.  The\r
   KDC implementation needs to consider expiration of tokens, key\r
   rollover and other security issues in token design.  The content of\r
   the cookie field is likely specific to the pre-authentication\r
   mechanisms used to authenticate the client.  If a client\r
   authentication response can be replayed to multiple KDCs via the\r
   PA_FX_COOKIE mechanism, an expiration in the cookie is RECOMMENDED to\r
   prevent the response being presented indefinitely.\r
\r
   If at least one more message for a mechanism or a mechanism set is\r
   expected by the KDC, the KDC returns a\r
   KDC_ERR_MORE_PREAUTH_DATA_NEEDED error with a PA_FX_COOKIE to\r
   identify the conversation with the client according to Section 6.5.4.\r
\r
        KDC_ERR_MORE_PREAUTH_DATA_NEEDED   TBA\r
\r
6.4.  Pre-authentication Set\r
\r
   If all mechanisms in a group need to successfully complete in order\r
   to authenticate a client, the client and the KDC SHOULD use the\r
   PA_AUTHENTICATION_SET padata element.\r
\r
   A PA_AUTHENTICATION_SET padata element contains the ASN.1 DER\r
   encoding of the PA-AUTHENTICATION-SET structure:\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 19]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
        PA-AUTHENTICATION-SET ::= SEQUENCE OF PA-AUTHENTICATION-SET-ELEM\r
\r
        PA-AUTHENTICATION-SET-ELEM ::= SEQUENCE {\r
            pa-type      [0] Int32,\r
                -- same as padata-type.\r
            pa-hint      [1] OCTET STRING OPTIONAL,\r
            pa-value  [2] OCTET STRING OPTIONAL,\r
            ...\r
        }\r
\r
   The pa-type field of the PA-AUTHENTICATION-SET-ELEM structure\r
   contains the corresponding value of padata-type in PA-DATA [RFC4120].\r
   Associated with the pa-type is a pa-hint, which is an octet-string\r
   specified by the pre-authentication mechanism.  This hint may provide\r
   information for the client which helps it determine whether the\r
   mechanism can be used.  For example a public-key mechanism might\r
   include the certificate authorities it trusts in the hint info.  Most\r
   mechanisms today do not specify hint info; if a mechanism does not\r
   specify hint info the KDC MUST NOT send a hint for that mechanism.\r
   To allow future revisions of mechanism specifications to add hint\r
   info, clients MUST ignore hint info received for mechanisms that the\r
   client believes do not support hint info.  The pa-value element of\r
   the PA-AUTHENTICATION-SET-ELEM sequence is included to carry the\r
   first padata-value from the KDC to the client.  If the client chooses\r
   this authentication set then the client MUST process this pa-value.\r
   The pa-value element MUST be absent for all but the first entry in\r
   the authentication set.  Clients MUST ignore pa-value for the second\r
   and following entries in the authentication set.\r
\r
   If the client chooses an authentication set, then its AS-REQ message\r
   MUST contain a PA_AUTHENTICATION_SET_SELECTED padata element.  This\r
   element contains the encoding of the PA-AUTHENTICATION-SET sequence\r
   received from the KDC corresponding to the authentication set that is\r
   chosen.  The client MUST use the same octet values received from the\r
   KDC; it cannot re-encode the sequence.  This allows KDCs to use bit-\r
   wise comparison to identify the selected authentication set.  The\r
   PA_AUTHENTICATION_SET_SELECTED padata element MUST come before any\r
   padata elements from the authentication set in the padata sequence in\r
   the AS-REQ message.  The client MAY cache authentication sets from\r
   prior messages and use them to construct an optimistic initial AS-\r
   REQ.  If the KDC receives a PA_AUTHENTICATION_SET_SELECTED padata\r
   element that does not correspond to an authentication set that it\r
   would offer, then the KDC returns the\r
   KDC_ERR_PREAUTH_BAD_AUTHENTICATION_SET error.  The edata in this\r
   error contains a sequence of padata just as for the\r
   KDC_ERR_PREAUTH_REQUIRED error.\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 20]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
      PA_AUTHENTICATION_SET_SELECTED         TBA\r
      KDC_ERR_PREAUTH_BAD_AUTHENTICATION_SET TBA\r
\r
   The PA-AUTHENTICATION-SET appears only in the first message from the\r
   KDC to the client.  In particular, the client MAY fail if the\r
   authentication mechanism sets change as the conversation progresses.\r
   Clients MAY assume that the hints provided in the authentication set\r
   contain enough information that the client knows what user interface\r
   elements need to be displayed during the entire authentication\r
   conversation.  Exceptional circumstances such as expired passwords or\r
   expired accounts may require that additional user interface be\r
   displayed.  Mechanism designers need to carefully consider the design\r
   of their hints so that the client has this information.  This way,\r
   clients can construct necessary dialogue boxes or wizards based on\r
   the authentication set and can present a coherent user interface.\r
   Current standards for user interface do not provide an acceptable\r
   experience when the client has to ask additional questions later in\r
   the conversation.\r
\r
   When indicating which sets of pre-authentication mechanisms are\r
   supported, the KDC includes a PA-AUTHENTICATION-SET padata element\r
   for each pre-authentication mechanism set.\r
\r
   The client sends the padata-value for the first mechanism it picks in\r
   the pre-authentication set, when the first mechanism completes, the\r
   client and the KDC will proceed with the second mechanism, and so on\r
   until all mechanisms complete successfully.  The PA_FX_COOKIE as\r
   defined in Section 6.3 MUST be sent by the KDC along with the first\r
   message that contains a PA-AUTHENTICATION-SET, in order to keep track\r
   of KDC states.\r
\r
   Before the authentication succeeds and a ticket is returned, the\r
   message that the client sends is an AS_REQ and the message that the\r
   KDC sends is a KRB-ERROR message.  The error code in the KRB-ERROR\r
   message from the KDC is KDC_ERR_MORE_PREAUTH_DATA_NEEDED as defined\r
   in Section 6.3 and the accompanying e-data contains the DER encoding\r
   of ASN.1 type METHOD-DATA.  The KDC includes the padata elements in\r
   the METHOD-DATA.  If there is no padata, the e-data field is absent\r
   in the KRB-ERROR message.\r
\r
   If the client sends the last message for a given mechanism, then the\r
   KDC sends the first message for the next mechanism.  If the next\r
   mechanism does not start with a KDC-side challenge, then the KDC\r
   includes a padata item with the appropriate pa-type and an empty pa-\r
   data.\r
\r
   If the KDC sends the last message for a particular mechanism, the KDC\r
   also includes the first padata for the next mechanism.\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 21]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
6.5.  Definition of Kerberos FAST Padata\r
\r
   As described in [RFC4120], Kerberos is vulnerable to offline\r
   dictionary attacks.  An attacker can request an AS-REP and try\r
   various passwords to see if they can decrypt the resulting ticket.\r
   RFC 4120 provides the entrypted timestap pre-authentication method\r
   that ameliorates the situation somewhat by requiring that an attacker\r
   observe a successful authentication.  However stronger security is\r
   desired in many environments.  The Kerberos FAST pre-authentication\r
   padata defined in this section provides a tool to significantly\r
   reduce vulnerability to offline dictionary attack.  When combined\r
   with encrypted timestamp, FAST requires an attacker to mount a\r
   successful man-in-the-middle attack to observe ciphertext.  When\r
   combined with host keys, FAST can even protect against active\r
   attacks.  FAST also provides solutions to common problems for pre-\r
   authentication mechanisms such as binding of the request and the\r
   reply, freshness guarantee of the authentication.  FAST itself,\r
   however, does not authenticate the client or the KDC, instead, it\r
   provides a typed hole to allow pre-authentication data be tunneled.\r
   A pre-authentication data element used within FAST is called a FAST\r
   factor.  A FAST factor captures the minimal work required for\r
   extending Kerberos to support a new pre-authentication scheme.\r
\r
   A FAST factor MUST NOT be used outside of FAST unless its\r
   specification explicitly allows so.  The typed holes in FAST messages\r
   can also be used as generic holes for other padata that are not\r
   intended to prove the client's identity, or establish the reply key.\r
\r
   New pre-authentication mechanisms SHOULD be designed as FAST factors,\r
   instead of full-blown pre-authentication mechanisms.\r
\r
   FAST factors that are pre-authentication mechanisms MUST meet the\r
   requirements in Section 5.\r
\r
   FAST employs an armoring scheme.  The armor can be a Ticket Granting\r
   Ticket (TGT) obtained by the client's machine using the host keys to\r
   pre-authenticate with the KDC, or an anonymous TGT obtained based on\r
   anonymous PKINIT [KRB-ANON] [RFC4556].\r
\r
   The rest of this section describes the types of armors and the syntax\r
   of the messages used by FAST.  Conforming implementations MUST\r
   support Kerberos FAST padata.\r
\r
   Any FAST armor scheme MUST provide a fresh armor key for each\r
   conversation.  Clients and KDCs can assume that if a message is\r
   encrypted and integrity protected with a given armor key then it is\r
   part of the conversation using that armor key.\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 22]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
6.5.1.  FAST Armors\r
\r
   An armor key is used to encrypt pre-authentication data in the FAST\r
   request and the response.  The KrbFastArmor structure is defined to\r
   identify the armor key.  This structure contains the following two\r
   fields: the armor-type identifies the type of armors, and the armor-\r
   value as an OCTET STRING contains the description of the armor scheme\r
   and the armor key.\r
\r
        KrbFastArmor ::= SEQUENCE {\r
            armor-type   [0] Int32,\r
                -- Type of the armor.\r
            armor-value  [1] OCTET STRING,\r
                -- Value of the armor.\r
            ...\r
        }\r
\r
   The value of the armor key is a matter of the armor type\r
   specification.  Only one armor type is defined in this document.\r
\r
        FX_FAST_ARMOR_AP_REQUEST           TBA\r
\r
   The FX_FAST_ARMOR_AP_REQUEST armor is based on Kerberos tickets.\r
\r
   Conforming implementations MUST implement the\r
   FX_FAST_ARMOR_AP_REQUEST armor type.\r
\r
6.5.1.1.  Ticket-based Armors\r
\r
   This is a ticket-based armoring scheme.  The armor-type is\r
   FX_FAST_ARMOR_AP_REQUEST, the armor-value contains an ASN.1 DER\r
   encoded AP-REQ.  The ticket in the AP-REQ is called an armor ticket\r
   or an armor TGT.  The subkey field in the AP-REQ MUST be present.\r
   The armor key is the subkey in the AP-REQ authenticator.\r
\r
   The server name field of the armor ticket MUST identify the TGS of\r
   the target realm.  Here are three ways in the decreasing preference\r
   order how an armor TGT SHOULD be obtained:\r
\r
   1.  If the client is authenticating from a host machine whose\r
       Kerberos realm has a trust path to the client's realm, the host\r
       machine obtains a TGT by pre-authenticating intitialy the realm\r
       of the host machine using the host keys.  If the client's realm\r
       is different than the realm of the local host, the machine then\r
       obtains a cross-realm TGT to the client's realm as the armor\r
       ticket.  Otherwise, the host's primary TGT is the armor ticket.\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 23]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   2.  If the client's host machine cannot obtain a host ticket strictly\r
       based on RFC4120, but the KDC has an asymmetric signing key that\r
       the client can verify the binding between the public key of the\r
       signing key and the expected KDC, the client can use anonymous\r
       PKINIT [KRB-ANON] [RFC4556] to authenticate the KDC and obtain an\r
       anonymous TGT as the armor ticket.  The armor key can be a cross-\r
       team TGT obtained based on the initial primary TGT obtained using\r
       anonymous PKINIT with KDC authentication.\r
\r
   3.  Otherwise, the client uses anonymous PKINIT to get an anonymous\r
       TGT without KDC authentication and that TGT is the armor ticket.\r
       Note that this mode of operation is vulnerable to man-in-the-\r
       middle attacks at the time of obtaining the initial anonymous\r
       armor TGT.  The armor key can be a cross-team TGT obtained based\r
       on the initial primary TGT obtained using anonymous PKINIT\r
       without KDC authentication.\r
\r
   Because the KDC does not know if the client is able to trust the\r
   ticket it has, the KDC MUST initialize the pre-authentication state\r
   to an unverified KDC.\r
\r
6.5.2.  FAST Request\r
\r
   A padata type PA_FX_FAST is defined for the Kerberos FAST pre-\r
   authentication padata.  The corresponding padata-value field\r
   [RFC4120] contains the DER encoding of the ASN.1 type PA-FX-FAST-\r
   REQUEST.\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 24]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
       PA_FX_FAST                         TBA\r
           -- Padata type for Kerberos FAST\r
\r
       PA-FX-FAST-REQUEST ::= CHOICE {\r
           armored-data [0] KrbFastArmoredReq,\r
           ...\r
       }\r
\r
       KrbFastArmoredReq ::= SEQUENCE {\r
           armor        [0] KrbFastArmor OPTIONAL,\r
               -- Contains the armor that identifies the armor key.\r
               -- MUST be present in AS-REQ.\r
               -- MUST be absent in TGS-REQ.\r
           req-checksum [1] Checksum,\r
               -- Checksum performed over the type KDC-REQ-BODY for\r
               -- the req-body field of the KDC-REQ structure defined in\r
               -- [RFC4120]\r
               -- The checksum key is the armor key, the checksum\r
               -- type is the required checksum type for the enctype of\r
               -- the armor key, and the key usage number is\r
               -- KEY_USAGE_FAST_REA_CHKSUM.\r
           enc-fast-req [2] EncryptedData, -- KrbFastReq --\r
               -- The encryption key is the armor key, and the key usage\r
               -- number is KEY_USAGE_FAST_ENC.\r
           ...\r
       }\r
\r
       KEY_USAGE_FAST_REA_CHKSUM          TBA\r
       KEY_USAGE_FAST_ENC                 TBA\r
\r
   The PA-FX-FAST-REQUEST structure contains a KrbFastArmoredReq type.\r
   The KrbFastArmoredReq encapsulates the encrypted padata.\r
\r
   The enc-fast-req field contains an encrypted KrbFastReq structure.\r
   The armor key is used to encrypt the KrbFastReq structure, and the\r
   key usage number for that encryption is KEY_USAGE_FAST_ARMOR.\r
\r
        KEY_USAGE_FAST_ARMOR               TBA\r
\r
   The armor key is selected as follows:\r
\r
   o  In an AS request, the armor field in the KrbFastArmoredReq\r
      structure MUST be present and the armor key is identified\r
      according to the specification of the armor type.\r
\r
   o  In a TGS request, the armor field in the KrbFastArmoredReq\r
      structure MUST NOT be present and the subkey in the AP-REQ\r
      authenticator in the PA-TGS-REQ PA-DATA MUST be present.  In this\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 25]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
      case, the armor key is that subkey in the AP-REQ authenticator.\r
\r
   The req-checksum field contains a checksum that is performed over the\r
   type KDC-REQ-BODY for the req-body field of the KDC-REQ [RFC4120]\r
   structure of the containing message.  The checksum key is the armor\r
   key, and the checksum type is the required checksum type for the\r
   enctype of the armor key per [RFC3961].  This checksum is included in\r
   order to bind the FAST data to the outer request.  A KDC that\r
   implements FAST will ignore the outer request, but including a\r
   checksum is relatively cheap and may prevent confusing behavior.\r
\r
   The KrbFastReq structure contains the following information:\r
\r
        KrbFastReq ::= SEQUENCE {\r
            fast-options [0] FastOptions,\r
                -- Additional options.\r
            padata       [1] SEQUENCE OF PA-DATA,\r
                -- padata typed holes.\r
            req-body     [2] KDC-REQ-BODY,\r
                -- Contains the KDC request body as defined in Section\r
                -- 5.4.1 of [RFC4120].\r
                -- This req-body field is preferred over the outer field\r
                -- in the KDC request.\r
             ...\r
        }\r
\r
   The fast-options field indicates various options that are to modify\r
   the behavior of the KDC.  The following options are defined:\r
\r
        FastOptions ::= KerberosFlags\r
            -- reserved(0),\r
            -- anonymous(1),\r
            -- kdc-referrals(16)\r
\r
\r
      Bits    Name          Description\r
     -----------------------------------------------------------------\r
      0     RESERVED        Reserved for future expansion of this field.\r
      1     anonymous       Requesting the KDC to hide client names in\r
                            the KDC response, as described next in this\r
                            section.\r
      16    kdc-referrals   Requesting the KDC to follow referrals, as\r
                            described next in this section.\r
\r
   Bits 1 through 15 (with bit 2 and bit 15 included) are critical\r
   options.  If the KDC does not support a critical option, it MUST fail\r
   the request with KDC_ERR_UNKNOWN_CRITICAL_FAST_OPTIONS (there is no\r
   accompanying e-data defined in this document for this error code).\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 26]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   Bit 16 and onward (with bit 16 included) are non-critical options.\r
   KDCs conforming to this specification ignores unknown non-critical\r
   options.\r
\r
        KDC_ERR_UNKNOWN_FAST_OPTIONS       TBA\r
\r
   The anonymous Option\r
\r
      The Kerberos response defined in [RFC4120] contains the client\r
      identity in clear text, This makes traffic analysis\r
      straightforward.  The anonymous option is designed to complicate\r
      traffic analysis.  If the anonymous option is set, the KDC\r
      implementing PA_FX_FAST MUST identify the client as the anonymous\r
      principal [KRB-ANON] in the KDC reply and the error response.\r
      Hence this option is set by the client if it wishes to conceal the\r
      client identity in the KDC response.  A conforming KD ignores the\r
      client principal name in the outer KDC-REQ-BODY field, and\r
      identifies the client using the cname and crealm fields in the\r
      req-body field of the KrbFastReq structure.\r
\r
   The kdc-referrals Option\r
\r
      The Kerberos client described in [RFC4120] has to request referral\r
      TGTs along the authentication path in order to get a service\r
      ticket for the target service.  The Kerberos client described in\r
      the [REFERRALS] need to contact the AS specified in the error\r
      response in order to complete client referrals.  The kdc-referrals\r
      option is designed to minimize the number of messages that need to\r
      be processed by the client.  This option is useful when, for\r
      example, the client may contact the KDC via a satellite link that\r
      has high network latency, or the client has limited computational\r
      capabilities.  If the kdc-referrals option is set, the KDC that\r
      honors this option acts as the client to follow AS referrals and\r
      TGS referrals [REFERRALS], and return the service ticket to the\r
      named server principal in the client request using the reply key\r
      expected by the client.  The kdc-referrals option can be\r
      implemented when the KDC knows the reply key.  The KDC can ignore\r
      kdc-referrals option when it does not understand it or it does not\r
      allow this option based on local policy.  The client SHOULD be\r
      able to process the KDC responses when this option is not honored\r
      by the KDC.\r
\r
   The padata field contains a list of PA-DATA structures as described\r
   in Section 5.2.7 of [RFC4120].  These PA-DATA structures can contain\r
   FAST factors.  They can also be used as generic typed-holes to\r
   contain data not intended for proving the client's identity or\r
   establishing a reply key, but for protocol extensibility.\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 27]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   The KDC-REQ-BODY in the FAST structure is used in preference to the\r
   KDC-REQ-BODY outside of the FAST pre-authentication.  The outer KDC-\r
   REQ-BODY structure SHOULD be filled in for backwards compatibility\r
   with KDCs that do not support FAST.  A conforming KDC ignores the\r
   outer KDC-REQ-BODY field in the KDC request.\r
\r
6.5.3.  FAST Response\r
\r
   The KDC that supports the PA_FX_FAST padata MUST include a PA_FX_FAST\r
   padata element in the KDC reply.  In the case of an error, the\r
   PA_FX_FAST padata is included in the KDC responses according to\r
   Section 6.5.4.\r
\r
   The corresponding padata-value field [RFC4120] for the PA_FX_FAST in\r
   the KDC response contains the DER encoding of the ASN.1 type PA-FX-\r
   FAST-REPLY.\r
\r
      PA-FX-FAST-REPLY ::= CHOICE {\r
          armored-data [0] KrbFastArmoredRep,\r
          ...\r
      }\r
\r
      KrbFastArmoredRep ::= SEQUENCE {\r
          enc-fast-rep      [0] EncryptedData, -- KrbFastResponse --\r
              -- The encryption key is the armor key in the request, and\r
              -- the key usage number is KEY_USAGE_FAST_REP.\r
          ...\r
      }\r
      KEY_USAGE_FAST_REP                 TBA\r
\r
   The PA-FX-FAST-REPLY structure contains a KrbFastArmoredRep\r
   structure.  The KrbFastArmoredRep structure encapsulates the padata\r
   in the KDC reply in the encrypted form.  The KrbFastResponse is\r
   encrypted with the armor key used in the corresponding request, and\r
   the key usage number is KEY_USAGE_FAST_REP.\r
\r
   The Kerberos client who does not receive a PA-FX-FAST-REPLY in the\r
   KDC response MUST support a local policy that rejects the response.\r
   Clients MAY also support policies that fall back to other mechanisms\r
   or that do not use pre-authentication when FAST is unavailable.  It\r
   is important to consider the potential downgrade attacks when\r
   deploying such a policy.\r
\r
   The KrbFastResponse structure contains the following information:\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 28]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
     KrbFastResponse ::= SEQUENCE {\r
         padata      [0] SEQUENCE OF PA-DATA,\r
             -- padata typed holes.\r
         rep-key     [1] EncryptionKey OPTIONAL,\r
             -- This, if present, replaces the reply key for AS and TGS.\r
             -- MUST be absent in KRB-ERROR.\r
         finished    [2] KrbFastFinished OPTIONAL,\r
             -- MUST be present if the client is authenticated,\r
             -- absent otherwise.\r
             -- Typically this is present if and only if the containing\r
             -- message is the last one in a conversation.\r
         ...\r
     }\r
\r
   The padata field in the KrbFastResponse structure contains a list of\r
   PA-DATA structures as described in Section 5.2.7 of [RFC4120].  These\r
   PA-DATA structures are used to carry data advancing the exchange\r
   specific for the FAST factors.  They can also be used as generic\r
   typed-holes for protocol extensibility.\r
\r
   The rep-key field, if present, contains the reply key that is used to\r
   encrypted the KDC reply.  The rep-key field MUST be absent in the\r
   case where an error occurs.  The enctype of the rep-key is the\r
   strongest mutually supported by the KDC and the client.\r
\r
   The finished field contains a KrbFastFinished structure.  It is\r
   filled by the KDC in the final message in the conversation; it MUST\r
   be absent otherwise.  In other words, this field can only be present\r
   in an AS-REP or a TGS-REP when a ticket is returned.\r
\r
   The KrbFastFinished structure contains the following information:\r
\r
        KrbFastFinished ::= SEQUENCE {\r
            timestamp   [0] KerberosTime,\r
            usec        [1] Microseconds,\r
                -- timestamp and usec represent the time on the KDC when\r
                -- the reply was generated.\r
            crealm      [2] Realm,\r
            cname       [3] PrincipalName,\r
                -- Contains the client realm and the client name.\r
            checksum    [4] Checksum,\r
                -- Checksum performed over all the messages in the\r
                -- conversation, except the containing message.\r
                -- The checksum key is the binding key as defined in\r
                -- Section 6.3, and the checksum type is the required\r
                -- checksum type of the binding key.\r
            ...\r
        }\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 29]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
        KEY_USAGE_FAST_FINISHED            TBA\r
\r
   The timestamp and usec fields represent the time on the KDC when the\r
   reply ticket was generated, these fields have the same semantics as\r
   the corresponding-identically-named fields in Section 5.6.1 of\r
   [RFC4120].  The client MUST use the KDC's time in these fields\r
   thereafter when using the returned ticket.  Note that the KDC's time\r
   in AS-REP may not match the authtime in the reply ticket if the kdc-\r
   referrals option is requested and honored by the KDC.\r
\r
   The cname and crealm fields identify the authenticated client.\r
\r
   The checksum field contains a checksum of all the messages in the\r
   conversation prior to the containing message (the containing message\r
   is excluded).  The checksum key is the binding key as defined in\r
   Section 6.3, and the checksum type is the required checksum type of\r
   the enctype of that key, and the key usage number is\r
   KEY_USAGE_FAST_FINISHED. [[anchor9: Examples would be good here; what\r
   all goes into the checksum?]]\r
\r
   When FAST padata is included, the PA-FX-COOKIE padata as defined in\r
   Section 6.3 MUST also be included if the KDC expects at least one\r
   more message from the client in order to complete the authentication.\r
\r
6.5.4.  Authenticated Kerberos Error Messages using Kerberos FAST\r
\r
   If the Kerberos FAST padata was included in the request, unless\r
   otherwise specified, the e-data field of the KRB-ERROR message\r
   [RFC4120] contains the ASN.1 DER encoding of the type METHOD-DATA\r
   [RFC4120] and a PA_FX_FAST is included in the METHOD-DATA.  The KDC\r
   MUST include all the padata elements such as PA-ETYPE-INFO2 and\r
   padata elments that indicate acceptable pre-authentication mechanisms\r
   [RFC4120] and in the KrbFastResponse structure.\r
\r
   If the Kerberos FAST padata is included in the request but not\r
   included in the error reply, it is a matter of the local policy on\r
   the client to accept the information in the error message without\r
   integrity protection.  The Kerberos client MAY process an error\r
   message without a PA-FX-FAST-REPLY, if that is only intended to\r
   return better error information to the application, typically for\r
   trouble-shooting purposes.\r
\r
   In the cases where the e-data field of the KRB-ERROR message is\r
   expected to carry a TYPED-DATA [RFC4120] element, the\r
   PA_FX_TYPED_DATA padata is included in the KrbFastResponse structure\r
   to encapsulate the TYPED-DATA [RFC4120] elements.  For example, the\r
   TD_TRUSTED_CERTIFIERS structure is expected to be in the KRB-ERROR\r
   message when the error code is KDC_ERR_CANT_VERIFY_CERTIFICATE\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 30]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   [RFC4556].\r
\r
        PA_FX_TYPED_DATA                   TBA\r
            -- This is the padata element that encapsulates a TYPED-DATA\r
            -- structure.\r
\r
   The corresponding padata-value for the PA_FX_TYPED_DATA padata type\r
   contains the DER encoding of the ASN.1 type TYPED-DATA [RFC4120].\r
\r
6.5.5.  The Encrypted Challenge FAST Factor\r
\r
   The encrypted challenge FAST factor authenticates a client using the\r
   client's long-term key.  This factor works similarly to the encrypted\r
   time stamp pre-authentication option described in [RFC4120].  The\r
   client encrypts a structure containing a timestamp in the challenge\r
   key.  The challenge key is KRB-FX-CF2(long_term_key, armor_key,\r
   "challengelongterm", "challengearmor").  Because the armor key is\r
   fresh and random, the challenge key is fresh and random.  The only\r
   purpose of the timestamp is to limit the validity of the\r
   authentication so that a request cannot be replayed.  A client MAY\r
   base the timestamp based on the KDC time in a KDC error and need not\r
   maintain accurate time synchronization itself.  If a client bases its\r
   time on an untrusted source, an attacker may trick the client into\r
   producing an authentication request that is valid at some future\r
   time.  The attacker may be able to use this authentication request to\r
   make it appear that a client has authenticated at that future time.\r
   If ticket-based armor is used, then the lifetime of the ticket will\r
   limit the window in which an attacker can make the client appear to\r
   have authenticated.  For many situations, the ability of an attacker\r
   to cause a client to appear to have authenticated is not a\r
   significant concern; the ability to avoid requiring time\r
   synchronization on clients is more valuable.\r
\r
   The client sends a padata of type PA_ENCRYPTED_CHALLENGE the\r
   corresponding padata-value contains the DER encoding of ASN.1 type\r
   EncryptedChallenge.\r
\r
      EncryptedChallenge ::= EncryptedData\r
              -- Encrypted PA-ENC-TS-ENC, encrypted in the challenge key\r
              -- using key usage KEY_USAGE_ENC_CHALLENGE_CLIENT for the\r
              --  client and KEY_USAGE_ENC_CHALLENGE_KDC for the KDC.\r
\r
      PA_ENCRYPTED_CHALLENGE          TBA\r
      KEY_USAGE_ENC_CHALLENGE_CLIENT  TBA\r
      KEY_USAGE_ENC_CHALLENGE_KDC     TBA\r
\r
   The client includes some time stamp reasonably close to the KDC's\r
   current time and encrypts it in the challenge key.  Clients MAY use\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 31]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   the current time; doing so prevents the exposure where an attacker\r
   can cause a client to appear to authenticate in the future.  The\r
   client sends the request including this factor.\r
\r
   On receiving an AS-REQ containing the PA_ENCRYPTED_CHALLENGE fast\r
   factor, the KDC decrypts the timestamp.  If the decryption fails the\r
   KDC SHOULD return KDC_ERR_PREAUTH_FAILED, including etype-info2 in\r
   the error [[anchor11: Or should this be KRB_APP_ERR_MODIFIED?]].  The\r
   KDC confirms that the timestamp falls within its current clock skew\r
   returning KRB_APP_ERR_SKEW if not.  The KDC then SHOULD check to see\r
   if the encrypted challenge is a replay.  The KDC MUST NOT consider\r
   two encrypted challenges replays simply because the time stamps are\r
   the same; to be a replay, the ciphertext MUST be identical.  It is\r
   not clear that RFC 3961 prevents encryption systems for which an\r
   attacker can transform one ciphertext into a different ciphertext\r
   yielding an identical plaintext.  So, it may not be safe to base\r
   replay detection on the ciphertext in the general case.  However the\r
   FAST tunnel provides integrity protection so requiring ciphertext be\r
   identical is secure in this instance.  Allowing clients to re-use\r
   time stamps avoids requiring that clients maintain state about which\r
   time stamps have been used.\r
\r
   If the KDC accepts the encrypted challenge, it MUST include a padata\r
   element of type PA_ENCRYPTED_CHALLENGE.  The KDC encrypts its current\r
   time in the challenge key.  The KDC MUST replace the reply key before\r
   issuing a ticket. [[anchor12: I'd like to say that the KDC replaces\r
   its reply key by this point.  However we need to decide at what\r
   points the FAST mechanism for replacing the reply key can be used and\r
   how that interacts with this.]]The client MUST check that the\r
   timestamp decrypts properly.  The client MAY check that the timestamp\r
   is in some reasonable skew of the current time.  The client MUST NOT\r
   require that the timestamp be identical to the timestamp in the\r
   issued credentials or the returned message.\r
\r
   The encrypted challenge FAST factor provides the following\r
   facilities: client-authentication, KDC authentication.  It does not\r
   provide the strengthening-reply-key facility.  The security\r
   considerations section of this document provides an explanation why\r
   the security requirements are met.\r
\r
   Conforming implementations MUST support the encrypted challenge FAST\r
   factor.\r
\r
6.6.  Authentication Strength Indication\r
\r
   Implementations that have pre-authentication mechanisms offering\r
   significantly different strengths of client authentication MAY choose\r
   to keep track of the strength of the authentication used as an input\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 32]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   into policy decisions.  For example, some principals might require\r
   strong pre-authentication, while less sensitive principals can use\r
   relatively weak forms of pre-authentication like encrypted timestamp.\r
\r
   An AuthorizationData data type AD-Authentication-Strength is defined\r
   for this purpose.\r
\r
        AD-authentication-strength         TBA\r
\r
   The corresponding ad-data field contains the DER encoding of the pre-\r
   authentication data set as defined in Section 6.4.  This set contains\r
   all the pre-authentication mechanisms that were used to authenticate\r
   the client.  If only one pre-authentication mechanism was used to\r
   authenticate the client, the pre-authentication set contains one\r
   element.\r
\r
   The AD-authentication-strength element MUST be included in the AD-IF-\r
   RELEVANT, thus it can be ignored if it is unknown to the receiver.\r
\r
\r
7.  IANA Considerations\r
\r
   This document defines several new pa-data types, key usages and error\r
   codes.  In addition it would be good to track which pa-data items are\r
   only to be used as FAST factors.\r
\r
\r
8.  Security Considerations\r
\r
   The kdc-referrals option in the Kerberos FAST padata requests the KDC\r
   to act as the client to follow referrals.  This can overload the KDC.\r
   To limit the damages of denied of service using this option, KDCs MAY\r
   restrict the number of simultaneous active requests with this option\r
   for any given client principal.\r
\r
   Because the client secrets are known only to the client and the KDC,\r
   the verification of the authenticated timestamp proves the client's\r
   identity, the verification of the authenticated timestamp in the KDC\r
   reply proves that the expected KDC responded.  The encrypted reply\r
   key is contained in the rep-key in the PA-FX-FAST-REPLY.  Therefore,\r
   the authenticated timestamp FAST factor as a pre-authentication\r
   mechanism offers the following facilities: client-authentication,\r
   replacing-reply-key, KDC-authentication.  There is no un-\r
   authenticated clear text introduced by the authenticated timestamp\r
   FAST factor.\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 33]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
9.  Acknowledgements\r
\r
   Sam Hartman would like to thank the MIT Kerberos Consortium for its\r
   funding of his time on this project prior to April 2008.\r
\r
   Several suggestions from Jeffery Hutzman based on early revisions of\r
   this documents led to significant improvements of this document.\r
\r
   The proposal to ask one KDC to chase down the referrals and return\r
   the final ticket is based on requirements in [ID.CROSS].\r
\r
   Joel Webber had a proposal for a mechanism similar to FAST that\r
   created a protected tunnel for Kerberos pre-authentication.\r
\r
\r
10.  References\r
\r
10.1.  Normative References\r
\r
   [KRB-ANON]\r
              Zhu, L. and P. Leach, "Kerberos Anonymity Support",\r
              draft-ietf-krb-wg-anon-04.txt (work in progress), 2007.\r
\r
   [REFERRALS]\r
              Raeburn, K. and L. Zhu, "Generating KDC Referrals to\r
              Locate Kerberos Realms",\r
              draft-ietf-krb-wg-kerberos-referrals-10.txt (work in\r
              progress), 2007.\r
\r
   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate\r
              Requirement Levels", BCP 14, RFC 2119, March 1997.\r
\r
   [RFC3961]  Raeburn, K., "Encryption and Checksum Specifications for\r
              Kerberos 5", RFC 3961, February 2005.\r
\r
   [RFC4120]  Neuman, C., Yu, T., Hartman, S., and K. Raeburn, "The\r
              Kerberos Network Authentication Service (V5)", RFC 4120,\r
              July 2005.\r
\r
   [RFC4556]  Zhu, L. and B. Tung, "Public Key Cryptography for Initial\r
              Authentication in Kerberos (PKINIT)", RFC 4556, June 2006.\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 34]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
   [SHA2]     National Institute of Standards and Technology, "Secure \r
              Hash Standard (SHS)", Federal Information Processing \r
              Standards Publication 180-2, August 2002.  \r
\r
   [X680]     ITU-T Recommendation X.680 (2002) | ISO/IEC 8824-1:2002,\r
              Information technology - Abstract Syntax Notation One\r
              (ASN.1): Specification of basic notation.\r
   \r
   [X690]     ITU-T Recommendation X.690 (2002) | ISO/IEC 8825-1:2002,\r
              Information technology - ASN.1 encoding Rules:\r
              Specification of Basic Encoding Rules (BER), Canonical\r
              Encoding Rules (CER) and Distinguished Encoding Rules\r
              (DER).              \r
\r
10.2.  Informative References\r
\r
   [EKE]      Bellovin, S. M. and M. Merritt. "Augmented \r
              Encrypted Key Exchange: A Password-Based Protocol Secure \r
              Against Dictionary Attacks and Password File Compromise". \r
              Proceedings of the 1st ACM Conference on Computer and \r
              Communications Security, ACM Press, November 1993.\r
   \r
   [HKDF]     Dang, Q. and P. Polk, draft-dang-nistkdf, work in \r
              progress.\r
\r
   [IEEE1363.2] \r
              IEEE P1363.2: Password-Based Public-Key Cryptography, \r
              2004.\r
\r
   [ID.CROSS]\r
              Sakane, S., Zrelli, S., and M. Ishiyama , "Problem\r
              Statement on the Operation of Kerberos in a Specific\r
              System", draft-sakane-krb-cross-problem-statement-02.txt\r
              (work in progress), April 2007.\r
\r
   [KRB-WG.SAM]\r
\r
              Hornstein, K., Renard, K., Neuman, C., and G. Zorn,\r
              "Integrating Single-use Authentication Mechanisms with\r
              Kerberos", draft-ietf-krb-wg-kerberos-sam-02.txt (work in\r
              progress), October 2003.\r
\r
\r
Appendix A.  Change History\r
\r
   RFC editor, please remove this section before publication.\r
\r
A.1.  Changes since 07\r
\r
      Propose replacement of authenticated timestamp with encrypted\r
      challenge.  The desire to avoid clients needing time\r
      synchronization and to simply the factor.\r
      Add a requirement that any FAST armor scheme must provide a fresh\r
      key for each conversation.  This allows us to assume that anything\r
      encrypted/integrity protected in the right key is fresh and not\r
      subject to cross-conversation cut&paste.\r
      Removed heartbeat padata.  The KDC will double up messages if it\r
      needs to; the client simply sends its message and waits for the\r
      next response.\r
      Define PA_AUTHENTICATION_SET_SELECTED\r
      Clarify a KDC cannot ignore padata is has clamed to support\r
\r
A.2.  Changes since 06\r
\r
      Note that even for replace reply key it is likely that the side\r
      using the mechanism will know that the other side supports it.\r
      Since it is reasonablly unlikely we'll need a container mechanism\r
      other than FAST itself, we don't need to optimize for that case.\r
      So, we want to optimize for implementation simplicity.  Thus if\r
      you do have such a container mechanism interacting with\r
      authentication sets we'll assume that the hint need to describe\r
      hints for all contained mechanisms.  This closes out a long-\r
      standing issue.\r
      Write up what Sam believes is the consensus on UI and prompts in\r
      the authentication set: clients MAY assume that they have all the\r
      UI information they need.\r
\r
\r
Appendix B.  ASN.1 module\r
\r
     KerberosPreauthFramework {\r
           iso(1) identified-organization(3) dod(6) internet(1)\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 35]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
           security(5) kerberosV5(2) modules(4) preauth-framework(3)\r
     } DEFINITIONS EXPLICIT TAGS ::= BEGIN\r
\r
     IMPORTS\r
          KerberosTime, PrincipalName, Realm, EncryptionKey, Checksum,\r
          Int32, EncryptedData, PA-ENC-TS-ENC, PA-DATA, KDC-REQ-BODY,\r
          Microseconds, KerberosFlags\r
               FROM KerberosV5Spec2 { iso(1) identified-organization(3)\r
                 dod(6) internet(1) security(5) kerberosV5(2)\r
                 modules(4) krb5spec2(2) };\r
                 -- as defined in RFC 4120.\r
\r
     PA-FX-COOKIE ::= SEQUENCE {\r
         conversationId  [0] OCTET STRING,\r
            -- Contains the identifier of this conversation. This field\r
            -- must contain the same value for all the messages\r
            -- within the same conversation.\r
         enc-binding-key [1] EncryptedData OPTIONAL,\r
                         -- EncryptionKey --\r
            -- This field is present when and only when a FAST\r
            -- padata as defined in Section 6.5 is included.\r
            -- The encrypted data, when decrypted, contains an\r
            -- EncryptionKey structure.\r
            -- This encryption key is encrypted using the armor key\r
            -- (defined in Section 6.5.1), and the key usage for the\r
            -- encryption is KEY_USAGE_FAST_BINDING_KEY.\r
            -- Present only once in a converstation.\r
         cookie          [2] OCTET STRING OPTIONAL,\r
            -- Opaque data, for use to associate all the messages in\r
            -- a single conversation between the client and the KDC.\r
            -- This is generated by the KDC and the client MUST copy\r
            -- the exact cookie encapsulated in a PA_FX_COOKIE data\r
            -- element into the next message of the same conversation.\r
         ...\r
     }\r
\r
     PA-AUTHENTICATION-SET ::= SEQUENCE OF PA-AUTHENTICATION-SET-ELEM\r
\r
     PA-AUTHENTICATION-SET-ELEM ::= SEQUENCE {\r
         pa-type      [0] Int32,\r
             -- same as padata-type.\r
         pa-hint      [1] OCTET STRING OPTIONAL,\r
         pa-value  [2] OCTET STRING OPTIONAL,\r
         ...\r
     }\r
\r
     KrbFastArmor ::= SEQUENCE {\r
         armor-type   [0] Int32,\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 36]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
             -- Type of the armor.\r
         armor-value  [1] OCTET STRING,\r
             -- Value of the armor.\r
         ...\r
     }\r
\r
     PA-FX-FAST-REQUEST ::= CHOICE {\r
         armored-data [0] KrbFastArmoredReq,\r
         ...\r
     }\r
\r
     KrbFastArmoredReq ::= SEQUENCE {\r
         armor        [0] KrbFastArmor OPTIONAL,\r
             -- Contains the armor that identifies the armor key.\r
             -- MUST be present in AS-REQ.\r
             -- MUST be absent in TGS-REQ.\r
         req-checksum [1] Checksum,\r
             -- Checksum performed over the type KDC-REQ-BODY for\r
             -- the req-body field of the KDC-REQ structure defined in\r
             -- [RFC4120]\r
             -- The checksum key is the armor key, the checksum\r
             -- type is the required checksum type for the enctype of\r
             -- the armor key, and the key usage number is\r
             -- KEY_USAGE_FAST_REA_CHKSUM.\r
         enc-fast-req [2] EncryptedData, -- KrbFastReq --\r
             -- The encryption key is the armor key, and the key usage\r
             -- number is KEY_USAGE_FAST_ENC.\r
         ...\r
     }\r
\r
     KrbFastReq ::= SEQUENCE {\r
         fast-options [0] FastOptions,\r
             -- Additional options.\r
         padata       [1] SEQUENCE OF PA-DATA,\r
             -- padata typed holes.\r
         req-body     [2] KDC-REQ-BODY,\r
             -- Contains the KDC request body as defined in Section\r
             -- 5.4.1 of [RFC4120].\r
             -- This req-body field is preferred over the outer field\r
             -- in the KDC request.\r
          ...\r
     }\r
\r
     FastOptions ::= KerberosFlags\r
         -- reserved(0),\r
         -- anonymous(1),\r
         -- kdc-referrals(16)\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 37]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
     PA-FX-FAST-REPLY ::= CHOICE {\r
         armored-data [0] KrbFastArmoredRep,\r
         ...\r
     }\r
\r
     KrbFastArmoredRep ::= SEQUENCE {\r
         enc-fast-rep      [0] EncryptedData, -- KrbFastResponse --\r
             -- The encryption key is the armor key in the request, and\r
             -- the key usage number is KEY_USAGE_FAST_REP.\r
         ...\r
     }\r
\r
     KrbFastResponse ::= SEQUENCE {\r
         padata      [0] SEQUENCE OF PA-DATA,\r
             -- padata typed holes.\r
         rep-key     [1] EncryptionKey OPTIONAL,\r
             -- This, if present, replaces the reply key for AS and TGS.\r
             -- MUST be absent in KRB-ERROR.\r
         finished    [2] KrbFastFinished OPTIONAL,\r
             -- MUST be present if the client is authenticated,\r
             -- absent otherwise.\r
             -- Typically this is present if and only if the containing\r
             -- message is the last one in a conversation.\r
         ...\r
     }\r
\r
     KrbFastFinished ::= SEQUENCE {\r
         timestamp   [0] KerberosTime,\r
         usec        [1] Microseconds,\r
             -- timestamp and usec represent the time on the KDC when\r
             -- the reply was generated.\r
         crealm      [2] Realm,\r
         cname       [3] PrincipalName,\r
             -- Contains the client realm and the client name.\r
         checksum    [4] Checksum,\r
             -- Checksum performed over all the messages in the\r
             -- conversation, except the containing message.\r
             -- The checksum key is the binding key as defined in\r
             -- Section 6.3, and the checksum type is the required\r
             -- checksum type of the binding key.\r
         ...\r
     }\r
\r
     EncryptedChallenge ::= EncryptedData\r
             -- Encrypted PA-ENC-TS-ENC, encrypted in the challenge key\r
             -- using key usage KEY_USAGE_ENC_CHALLENGE_CLIENT for the\r
             --  client and KEY_USAGE_ENC_CHALLENGE_KDC for the KDC.\r
     END\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 38]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
Authors' Addresses\r
\r
   Larry Zhu\r
   Microsoft Corporation\r
   One Microsoft Way\r
   Redmond, WA  98052\r
   US\r
\r
   Email: lzhu@microsoft.com\r
\r
\r
   Sam hartman\r
   Painless Security\r
\r
   Email: hartmans-ietf@mit.edu\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 39]\r
\f\r
Internet-Draft         Kerberos Preauth Framework              July 2008\r
\r
\r
Full Copyright Statement\r
\r
   Copyright (C) The IETF Trust (2008).\r
\r
   This document is subject to the rights, licenses and restrictions\r
   contained in BCP 78, and except as set forth therein, the authors\r
   retain all their rights.\r
\r
   This document and the information contained herein are provided on an\r
   "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS\r
   OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND\r
   THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS\r
   OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF\r
   THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED\r
   WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.\r
\r
\r
Intellectual Property\r
\r
   The IETF takes no position regarding the validity or scope of any\r
   Intellectual Property Rights or other rights that might be claimed to\r
   pertain to the implementation or use of the technology described in\r
   this document or the extent to which any license under such rights\r
   might or might not be available; nor does it represent that it has\r
   made any independent effort to identify any such rights.  Information\r
   on the procedures with respect to rights in RFC documents can be\r
   found in BCP 78 and BCP 79.\r
\r
   Copies of IPR disclosures made to the IETF Secretariat and any\r
   assurances of licenses to be made available, or the result of an\r
   attempt made to obtain a general license or permission for the use of\r
   such proprietary rights by implementers or users of this\r
   specification can be obtained from the IETF on-line IPR repository at\r
   http://www.ietf.org/ipr.\r
\r
   The IETF invites any interested party to bring to its attention any\r
   copyrights, patents or patent applications, or other proprietary\r
   rights that may cover technology that may be required to implement\r
   this standard.  Please address the information to the IETF at\r
   ietf-ipr@ietf.org.\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
\r
Zhu & Hartman           Expires January 15, 2009               [Page 40]\r
\f\r
\r