Independent Submission                                          M. Saito
Request for Comments: 6193                            NTT Communications
Category: Informational                                          D. Wing
ISSN: 2070-1721                                            Cisco Systems
                                                               M. Toyama
                                                         NTT Corporation
                                                              April 2011

Media Description for the Internet Key Exchange Protocol (IKE)
in the Session Description Protocol (SDP)

Abstract

   This document specifies how to establish a media session that
   represents a virtual private network using the Session Initiation
   Protocol for the purpose of on-demand media/application sharing
   between peers.  It extends the protocol identifier of the Session
   Description Protocol (SDP) so that it can negotiate use of the
   Internet Key Exchange Protocol (IKE) for media sessions in the SDP
   offer/answer model.  It also specifies a method to boot up IKE and
   generate IPsec security associations using a self-signed certificate.

Status of This Memo

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

   This is a contribution to the RFC Series, independently of any other
   RFC stream.  The RFC Editor has chosen to publish this document at
   its discretion and makes no statement about its value for
   implementation or deployment.  Documents approved for publication by
   the RFC Editor are not a candidate for any level of Internet
   Standard; see Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc6193.

Copyright Notice

   Copyright (c) 2011 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.

   1. Applicability Statement .........................................3
   2. Introduction ....................................................3
      2.1. Problem Statement ..........................................4
      2.2. Approach to Solution .......................................4
      2.3. Alternative Solution under Prior Relationship
           between Two Nodes ..........................................6
      2.4. Authorization Model ........................................6
      2.5. Conventions Used in This Document ..........................6
   3. Protocol Overview ...............................................7
   4. Protocol Identifiers ............................................8
   5. Normative Behavior ..............................................9
      5.1. SDP Offer and Answer Exchange ..............................9
      5.2. Maintenance and Termination of VPN Session ................10
      5.3. Forking ...................................................11
      5.4. Port Usage ................................................11
      5.5. Multiplexing UDP Messages When Using ICE ..................11
   6. Examples .......................................................13
      6.1. Example of SDP Offer and Answer Exchange without
           IPsec NAT-Traversal .......................................13
      6.2. Example of SDP Offer and Answer Exchange with
           IPsec NAT-Traversal .......................................14
   7. Application to IKE .............................................15
   8. Specifications Assuming Prior Relationship between Two Nodes ...16
      8.1. Certificates Signed by Trusted Third Party ................16
      8.2. Configured Pre-Shared Key .................................16
   9. Security Considerations ........................................17
   10. IANA Considerations ...........................................19
   11. Acknowledgments ...............................................20
   12. References ....................................................20
      12.1. Normative References .....................................20
      12.2. Informative References ...................................21

1. Applicability Statement

   This document provides information about a deployed use of the
   Session Initiation Protocol (SIP) [RFC3261] for the Internet
   community.  It is not currently an IETF standards track proposal.
   The mechanisms in this document use SIP as a name resolution and
   authentication mechanism to initiate an Internet Key Exchange
   Protocol (IKE) [RFC5996] session.  The purpose of this document is to
   establish an on-demand virtual private network (VPN) to a home router
   that does not have a fixed IP address using self-signed certificates.
   It is only applicable under the condition that the integrity of the
   Session Description Protocol (SDP) [RFC4566] is assured.  The method
   to ensure this integrity of SDP is outside the scope of this
   document.  This document specifies the process in which a pair of SIP
   user agents resolve each other's names, exchange the fingerprints of
   their self-signed certificates securely, and agree to establish an
   IPsec-based VPN [RFC4301].  However, this document does not make any
   modifications to the specifications of IPsec/IKE.  Despite the
   limitations of the conditions under which this document can be
   applied, there are sufficient use cases in which this specification
   is helpful, such as the following:

   o  Sharing media using a framework developed by Digital Living
      Network Alliance (DLNA) or similar protocols over VPN between two
      user devices.

   o  Accessing remote desktop applications over VPN initiated by SIP
      call.  As an additional function of click-to-call, a customer
      service agent can access a customer's PC remotely to troubleshoot
      the problem while talking with the customer over the phone.

   o  Accessing and controlling medical equipment (medical robotics)
      remotely to monitor the elderly in a rural area (remote care
      services).

   o  Using a LAN-based gaming protocol based on peer-to-peer rather
      than via a gaming server.

2. Introduction

This section describes the problem in accessing home networks and
provides an overview of the proposed solution.

2.1. Problem Statement

   Home servers and network-capable consumer electronic devices have
   been widely deployed.  People using such devices are willing to share
   content and applications and are therefore seeking ways to establish
   multiple communication channels with each other.  However, there are
   several obstacles to be overcome in the case of remote home access.

   It is often not possible for a device outside the home network to
   connect to another device inside the home network because the home
   device is behind a network address translation (NAT) or firewall that
   allows outgoing connections but blocks incoming connections.  One
   effective solution for this problem is VPN remote access to the NAT
   device, which is usually a home router.  With this approach, once the
   external device joins the home network securely, establishing
   connections with all the devices inside the home will become easy
   because popular LAN-based communication methods such as DLNA can be
   used transparently.  However, there are more difficult cases in which
   a home router itself is located behind the NAT.  In such cases, it is
   also necessary to consider NAT traversal of the remote access to the
   home router.  In many cases, because the global IP address of the
   home router is not always fixed, it is necessary to make use of an
   effective name resolution mechanism.

   In addition, there is the problem of how a remote client and a home
   router authenticate each other over IKE to establish IPsec for remote
   access.  It is not always possible for the two devices to securely
   exchange a pre-shared key in advance.  Administrative costs can make
   it impractical to distribute authentication certificates signed by a
   well-known root certification authority (CA) to all the devices.  In
   addition, it is inefficient to publish a temporary certificate to a
   device that does not have a fixed IP address or hostname.  To resolve
   these authentication issues, this document proposes a mechanism that
   enables the devices to authenticate each other using self-signed
   certificates.

2.2. Approach to Solution

   This document proposes the use of SIP as a name resolution and
   authentication mechanism because of three main advantages:

   o  Delegation of Authentication to Third Party

      Devices can be free from managing their signed certificates and
      whitelists by taking advantage of authentication and authorization
      mechanisms supported by SIP.

   o  UDP Hole Punching for IKE/IPsec

      SIP has a cross-NAT rendezvous mechanism, and Interactive
      Connectivity Establishment (ICE) [RFC5245] has a function to open
      ports through the NAT.  The combination of these effective
      functions can be used for general applications as well as real-
      time media.  It is difficult to set up a session between devices
      without SIP if the devices are behind various types of NAT.

   o  Reuse of Existing SIP Infrastructure

      SIP servers are widely distributed as a scalable infrastructure,
      and it is quite practical to reuse them without any modifications.

   Today, SIP is applied to not only Voice over IP (VoIP) but also
   various applications and is recognized as a general protocol for
   session initiation.  Therefore, it can also be used to initiate
   IKE/IPsec sessions.

   However, there is also a specification that uses a self-signed
   certificate for authentication in the SIP/SDP framework.
   "Connection-Oriented Media Transport over the Transport Layer
   Security (TLS) Protocol in the Session Description Protocol (SDP)"
   [RFC4572] (hereafter referred to as comedia-tls) specifies a method
   to exchange the fingerprint of a self-signed certificate to establish
   a Transport Layer Security (TLS) [RFC5246] connection.  This
   specification defines a mechanism by which self-signed certificates
   can be used securely, provided that the integrity of the SDP
   description is assured.  Because a certificate itself is used for
   authentication not only in TLS but also in IKE, this mechanism will
   be applied to the establishment of an IPsec security association (SA)
   by extending the protocol identifier of SDP so that it can specify
   IKE.

   One easy method to protect the integrity of the SDP description,
   which is the premise of this specification, is to use the SIP
   identity [RFC4474] mechanism.  This approach is also referred to in
   [RFC5763].  Because the SIP identity mechanism can protect the
   integrity of a body part as well as the value of the From header in a
   SIP request by using a valid Identity header, the receiver of the
   request can establish secure IPsec connections with the sender by
   confirming that the hash value of the certificate sent during IKE
   negotiation matches the fingerprint in the SDP.  Although SIP
   identity does not protect the identity of the receiver of the SIP
   request, SIP-connected identity [RFC4916] does.  Note that the
   possible deficiencies discussed in [RFC4474-Concerns] could affect
   this specification if SIP identity is used for the security
   mechanism.

   Considering the above background, this document defines new media
   formats "ike-esp" and "ike-esp-udpencap", which can be used when the
   protocol identifier is "udp", to enable the negotiation of using IKE
   for media sessions over SDP exchange on the condition that the
   integrity of the SDP description is assured.  It also specifies the
   method to set up an IPsec SA by exchanging fingerprints of self-
   signed certificates based on comedia-tls, and it notes the example of
   SDP offer/answer [RFC3264] and the points that should be taken care
   of by implementation.  Because there is a chance that devices are
   behind NAT, this document also covers the method to combine IKE/IPsec
   NAT-Traversal [RFC3947][RFC3948] with ICE.  In addition, it defines
   the attribute "ike-setup" for IKE media sessions, similar to the
   "setup" attribute for TCP-based media transport defined in RFC 4145
   [RFC4145].  This attribute is used to negotiate the role of each
   endpoint in the IKE session.

2.3. Alternative Solution under Prior Relationship between Two Nodes

   Under quite limited conditions, certificates signed by trusted third
   parties or pre-shared keys between endpoints could be used for
   authentication in IKE, using SIP servers only for name resolution and
   authorization of session initiation.  Such limited cases are
   addressed in Section 8.

2.4. Authorization Model

   In this document, SIP servers are used for authorization of each SIP
   call.  The actual media sessions of IPsec/IKE are not authorized by
   SIP servers but by the remote client and the home router based on the
   information in SIP/SDP.  For example, the home router recognizes the
   remote client with its SIP-URI and IP address in the SDP.  If it
   decides to accept the remote client as a peer of a VPN session, it
   will accept the following IKE session.  Then, during the IKE
   negotiation, the certificate fingerprint in the SDP is compared with
   the certificate exchanged in the IKE session.  If they match, IKE
   negotiation continues.  Only a successful IKE negotiation establishes
   an IPsec session with the remote peer.

2.5. Conventions Used in This Document

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
   document are to be interpreted as described in [RFC2119].

3. Protocol Overview

   Figure 1 shows a case of VPN remote access from a device outside the
   home to a home router whose IP address is not fixed.  In this case,
   the external device, a remote client, recognizes the Address of
   Record of the home router but does not have any information about its
   contact address and certificate.  Generally, establishing an IPsec SA
   dynamically and securely in this situation is difficult.  However, as
   specified in comedia-tls [RFC4572], if the integrity of SDP session
   descriptions is assured, it is possible for the home router and the
   remote client to have a prior relationship with each other by
   exchanging certificate fingerprints, i.e., secure one-way hashes of
   the distinguished encoding rules (DER) form of the certificates.

              REGISTRATION                REGISTRATION
                 (1)       +----------+      (1)
            +------------->|          |<---------+
            |    INVITE(2) |          |          |
            | +----------->|   SIP    |--------+ |
            | |  200 OK(2) |   Proxy  |        | |
            | | +----------|          |<-----+ | |
            | | |          |          |      | | |  _________
            | | V          +----------+      | V | /         \
         +----------+ IKE (Media Session) +---------+         \
         | Remote   |<---------(3)------->| Home    |  Home    \
         | Client   |                     | Router  | Network   |
         |         ============(4)====================          |
         |(SIP UAC) |     VPN (IPsec SA)  |(SIP UAS)|          /
         +----------+                     +---------+         /
                                                   \_________/

               Figure 1: Remote Access to Home Network

   (1)  Both Remote Client and Home Router generate secure signaling
        channels.  They may REGISTER to SIP Proxy using TLS.

   (2)  Remote Client sends an offer SDP with an INVITE request to Home
        Router, and Home Router returns an answer SDP with a reliable
        response (e.g., 200 OK).  Both exchange the fingerprints of
        their self-signed certificates in SDP during this transaction.
        Remote Client does not accept an answer SDP with an unreliable
        response as the final response.

   (3)  After the SDP exchange, Remote Client, which has the active
        role, initiates IKE with Home Router, which has the passive
        role, to establish an IPsec SA.  Both validate that the
        certificate presented in the IKE exchange has a fingerprint that

        matches the fingerprint from SDP.  If they match, IKE
        negotiation proceeds as normal.

   (4)  Remote Client joins the Home Network.

   By this method, the self-signed certificates of both parties are used
   for authentication in IKE, but SDP itself is not concerned with all
   the negotiations related to key-exchange, such as those of encryption
   and authentication algorithms.  These negotiations are up to IKE.  In
   many cases where IPsec is used for remote access, a remote client
   needs to dynamically obtain a private address inside the home network
   while initiating the remote access.  Therefore, the IPsec security
   policy also needs to be set dynamically at the same time.  However,
   such a management function of the security policy is the
   responsibility of the high-level application.  SDP is not concerned
   with it.  The roles of SDP here are to determine the IP addresses of
   both parties used for IKE connection with c-line in SDP and to
   exchange the fingerprints of the certificates used for authentication
   in IKE with the fingerprint attribute in SDP.

4. Protocol Identifiers

   This document defines two SDP media formats for the "udp" protocol
   under the "application" media type: "ike-esp" and "ike-esp-udpencap".
   The format "ike-esp" indicates that the media described is IKE for
   the establishment of an IPsec security association as described in
   IPsec Encapsulating Security Payload (ESP) [RFC4303].  In contrast,
   "ike-esp-udpencap" indicates that the media described is IKE, which
   is capable of NAT traversal for the establishment of UDP
   encapsulation of IPsec packets through NAT boxes as specified in
   [RFC3947] and [RFC3948].  Even if the offerer and answerer exchange
   "ike-esp-udpencap", IKE conforming to [RFC3947] and [RFC3948] can end
   up establishing a normal IPsec tunnel when there is no need to use
   UDP encapsulation of IPsec.  Both the offerer and answerer can
   negotiate IKE by specifying "udp" in the "proto" field and "ike-esp"
   or "ike-esp-udpencap" in the "fmt" field in SDP.

   In addition, this document defines a new attribute "ike-setup", which
   can be used when the protocol identifier is "udp" and the "fmt" field
   is "ike-esp" or "ike-esp-udpencap", in order to describe how
   endpoints should perform the IKE session setup procedure.  The "ike-
   setup" attribute indicates which of the end points should initiate
   the establishment of an IKE session.  The "ike-setup" attribute is
   charset-independent and can be a session- or media-level attribute.
   The following is the ABNF of the "ike-setup" attribute.

      ike-setup-attr = "a=ike-setup:" role
      role           = "active" / "passive" / "actpass"

      'active':   The endpoint will initiate an outgoing session.
      'passive':  The endpoint will accept an incoming session.
      'actpass':  The endpoint is willing to accept an incoming
                  session or to initiate an outgoing session.

   Both endpoints use the SDP offer/answer model to negotiate the value
   of "ike-setup", following the procedures determined for the "setup"
   attribute defined in Section 4.1 of [RFC4145].  However, "holdconn",
   as defined in [RFC4145], is not defined for the "ike-setup"
   attribute.

      Offer       Answer
      ----------------------------
      active      passive
      passive     active
      actpass     active / passive

   The semantics for the "ike-setup" attribute values of "active",
   "passive", and "actpass" in the offer/answer exchange are the same as
   those described for the "setup" attribute in Section 4.1 of
   [RFC4145], except that "ike-setup" applies to an IKE session instead
   of a TCP connection.  The default value of the "ike-setup" attribute
   is "active" in the offer and "passive" in the answer.

5. Normative Behavior

In this section, a method to negotiate the use of IKE for media
sessions in the SDP offer/answer model is described.

5.1. SDP Offer and Answer Exchange

   An offerer and an answerer negotiate the use of IKE following the
   usage of the protocol identifiers defined in Section 4.  If IPsec
   NAT-Traversal is not necessary, the offerer MAY use the media format
   "ike-esp" to indicate an IKE session.

   If either of the endpoints that negotiate IKE is behind the NAT, the
   endpoints need to transmit both IKE and IPsec packets over the NAT.
   That mechanism is specified in [RFC3947] and [RFC3948]: both
   endpoints encapsulate IPsec-ESP packets with a UDP header and
   multiplex them into the UDP path that IKE generates.

   To indicate this type of IKE session, the offerer uses "ike-esp-
   udpencap" media lines.  In this case, the offerer MAY decide their
   transport addresses (combination of IP address and port) before

   starting IKE, making use of the ICE framework.  Because UDP-
   encapsulated ESP packets and IKE packets go through the same UDP hole
   of a NAT, IPsec NAT-Traversal works if ICE reserves simply one UDP
   path through the NAT.  However, those UDP packets need to be
   multiplexed with Session Traversal Utilities for NAT (STUN) [RFC5389]
   packets if ICE is required to use STUN.  A method to coordinate IPsec
   NAT-Traversal and ICE is described in Sections 5.4 and 5.5.

   The offer MAY contain media lines for media other than "ike-esp" or
   "ike-esp-udpencap".  For example, audio stream may be included in the
   same SDP to have a voice session when establishing the VPN.  This may
   be useful to verify that the connected device is indeed operated by
   somebody who is authorized to access it, as described in Section 9.
   If that occurs, the negotiation described in this specification
   occurs only for the "ike-esp" or "ike-esp-udpencap" media lines;
   other media lines are negotiated and set up normally.  If the
   answerer determines it will refuse the IKE session without beginning
   the IKE negotiation (e.g., the From address is not on the permitted
   list), it SHOULD reject the "ike-esp" or "ike-esp-udpencap" media
   line in the normal manner by setting the port number in the SDP
   answer to 0 and SHOULD process the other media lines normally (only
   if it is still reasonable to establish that media without VPN).

   If the offerer and the answerer agree to start an IKE session by the
   offer/answer exchange, they will start the IKE setup.  Following the
   comedia-tls specification [RFC4572], the fingerprint attribute, which
   may be either a session- or a media-level SDP attribute, is used to
   exchange fingerprints of self-signed certificates.  If the
   fingerprint attribute is a session-level attribute, it applies to all
   IKE sessions and TLS sessions for which no media-level fingerprint
   attribute is defined.

   Note that it is possible for an offerer to become the IKE responder
   and an answerer to become the IKE initiator.  For example, when a
   Remote Access Server (RAS) sends an INVITE to an RAS client, the
   server may expect the client to become an IKE initiator.  In this
   case, the server sends an offer SDP with ike-setup:passive and the
   client returns an answer SDP with ike-setup:active.

5.2. Maintenance and Termination of VPN Session

   If the high-level application recognizes a VPN session as the media
   session, it MAY discard the IPsec SA and terminate IKE when that
   media session is terminated by a BYE request.  Therefore, the
   application aware of the VPN session MUST NOT send a BYE request as
   long as it needs the IPsec SA.  On the other hand, if the high-level
   application detects that a VPN session is terminated, it MAY
   terminate the media associated with the VPN or the entire SIP

   session.  Session timers in SIP [RFC4028] MAY be used for the session
   maintenance of the SIP call, but this does not necessarily ensure
   that the VPN session is alive.  If the VPN session needs session
   maintenance such as keep-alive and rekeying, it MUST be done
   utilizing its own maintenance mechanisms.  SIP re-INVITE MUST NOT be
   used for this purpose.  Note that each party can cache the
   certificate of the other party as described in the Security
   Considerations section of comedia-tls [RFC4572].

5.3. Forking

   Forking to multiple registered instances is outside the scope of this
   document.  At least, it is assumed that a User Agent Client (UAC)
   establishes a session with only one User Agent Server (UAS).
   Encountering forked answers should be treated as an illegal process,
   and the UAC should cancel the session.

5.4. Port Usage

   IKE generally uses local UDP port 500, but the IPsec NAT-Traversal
   specification requires a port transition to local UDP port 4500
   during IKE negotiation because IPsec-aware NAT may multiplex IKE
   sessions using port 500 without changing the port number.  If using
   ICE for IPsec Nat-Traversal, this port transition of IKE means ICE
   has to generate an additional UDP path for port 4500, and this would
   be unnecessary overhead.  However, IPsec NAT-Traversal allows an IKE
   session to use local UDP port 4500 from the beginning without using
   port 500.  Therefore, the endpoints SHOULD use their local UDP port
   4500 for an IKE session from the beginning, and ICE will only need to
   generate a UDP path of port 4500.

   When using ICE, a responder's IKE port observed by an initiator is
   not necessarily 500 or 4500.  Therefore, an IKE initiator MUST allow
   any destination ports in addition to 500 and 4500 for the IKE packets
   that it sends.  An IKE initiator just initiates an IKE session to the
   port number decided by an SDP offer/answer or ICE.

5.5. Multiplexing UDP Messages When Using ICE

   Conforming to ICE, an offerer and an answerer start a STUN
   connectivity check after SDP exchange.  Then the offerer initiates
   the IKE session making use of the UDP path generated by STUN packets.
   In addition, UDP-encapsulated ESP packets are multiplexed into the
   same UDP path as IKE.  Thus, it is necessary to multiplex the three
   different packets, STUN, IKE, and UDP-encapsulated ESP, into the same
   UDP path.  This section describes how to demultiplex these three
   packets.

   At the first step, the endpoint that received a UDP packet at the
   multiplexed port MUST check the first 32 bits (bits 0-31) of the UDP
   payload.  If they are all 0, which is defined as a non-ESP marker,
   that packet MUST be treated as an IKE packet.

   Otherwise, it is judged as an ESP packet in the IPsec NAT-Traversal
   specification.  It is furthermore necessary to distinguish STUN from
   ESP.  Therefore, the bits 32-63 from the beginning of the UDP payload
   MUST be checked.  If the bits do not match the magic cookie of STUN
   0x2112A442 (most packets do not match), the packet is treated as an
   ESP packet because it is no longer a STUN packet.

   However, if the bits do match the magic cookie, an additional test is
   necessary to determine if the packet is STUN or ESP.  The magic
   cookie field of STUN overlaps the sequence number field of ESP, so a
   possibility still remains that the sequence number of ESP coincides
   with 0x2112A442.  In this additional test, the validity of the
   fingerprint attribute of the STUN message MUST be checked.  If there
   is a valid fingerprint in the message, it is judged as a STUN packet;
   otherwise, it is an ESP packet.

   The above logic is expressed as follows.

      if SPI-field-is-all-zeros
           { packet is IKE }
        else
           {
           if bits-32-through-63 == stun-magic-cookie-value and
              bits-0-through-1 == 0 and
              bits-2-through-15 == a STUN message type and
              bits-16-through-31 == length of this UDP packet
              {
               fingerprint_found == parse_for_stun_fingerprint();
               if fingerprint_found == 1
                  { packet is STUN }
               else
                  { packet is ESP }
              }
           else
              { packet is ESP }
           }

6. Examples

6.1. Example of SDP Offer and Answer Exchange without IPsec NAT-
Traversal

   If IPsec NAT-Traversal is not necessary, SDP negotiation to set up
   IKE is quite simple.  Examples of SDP exchange are as follows.

   (Note: Due to RFC formatting conventions, this document splits SDP
   across lines whose content would exceed 72 characters.  A backslash
   character marks where this line folding has taken place.  This
   backslash and its trailing CRLF and whitespace would not appear in
   actual SDP content.)

   offer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.10
      a=ike-setup:active
      a=fingerprint:SHA-1 \
      4A:AD:B9:B1:3F:82:18:3B:54:02:12:DF:3E:5D:49:6B:19:E5:7C:AB
      ...

   answer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.20
      a=ike-setup:passive
      a=fingerprint:SHA-1 \
      D2:9F:6F:1E:CD:D3:09:E8:70:65:1A:51:7C:9D:30:4F:21:E4:4A:8E
      ...

      Figure 2: SDP Example When Offerer Is an IKE Initiator

   offer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.10
      a=ike-setup:passive
      a=fingerprint:SHA-1 \
      4A:AD:B9:B1:3F:82:18:3B:54:02:12:DF:3E:5D:49:6B:19:E5:7C:AB
      ...

   answer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.20
      a=ike-setup:active
      a=fingerprint:SHA-1 \
      D2:9F:6F:1E:CD:D3:09:E8:70:65:1A:51:7C:9D:30:4F:21:E4:4A:8E
      ...

      Figure 3: SDP Example When Offerer Is an IKE Responder

6.2. Example of SDP Offer and Answer Exchange with IPsec NAT-Traversal

   We consider the following scenario here.

                      +---------------------+
                      |                     |
                      |      Internet       |
                      |                     |
                      +---------------------+
                        |                |
                        |                |(192.0.2.20:45664)
                        |           +---------+
                        |           |   NAT   |
                        |           +---------+
                        |                |
       (192.0.2.10:4500)|                |(192.0.2.100:4500)
                   +---------+      +----------+
                   | offerer |      | answerer |
                   +---------+      +----------+

                  Figure 4: NAT-Traversal Scenario

   As shown above, an offerer is on the Internet, but an answerer is
   behind the NAT.  The offerer cannot initiate an IKE session unless
   the answerer prepares a global routable transport address that
   accepts IKE packets.  In this case, the following offer/answer
   exchange will take place.

   offer SDP
      ...
      a=ice-pwd:YH75Fviy6338Vbrhrlp8Yh
      a=ice-ufrag:9uB6
      m=application 4500 udp ike-esp-udpencap
      c=IN IP4 192.0.2.10
      a=ike-setup:active
      a=fingerprint:SHA-1 \
      4A:AD:B9:B1:3F:82:18:3B:54:02:12:DF:3E:5D:49:6B:19:E5:7C:AB
      a=candidate:1 1 udp 2130706431 192.0.2.10 4500 typ host
      ...

   answer SDP
      ...
      a=ice-pwd:asd88fgpdd777uzjYhagZg
      a=ice-ufrag:8hhY
      m=application 45664 udp ike-esp-udpencap
      c=IN IP4 192.0.2.20
      a=ike-setup:passive
      a=fingerprint:SHA-1 \
      D2:9F:6F:1E:CD:D3:09:E8:70:65:1A:51:7C:9D:30:4F:21:E4:4A:8E
      a=candidate:1 1 udp 2130706431 192.0.2.100 4500 typ host
      a=candidate:2 1 udp 1694498815 192.0.2.20 45664 typ srflx \
      raddr 192.0.2.100 rport 4500
      ...

      Figure 5: SDP Example with IPsec NAT-Traversal

7. Application to IKE

   After the fingerprints of both parties are securely shared over the
   SDP exchange, the IKE initiator MAY start the IKE session with the
   other party.  To follow this specification, a digital signature MUST
   be chosen as an authentication method in IKE phase 1.  In this
   process, a certificate whose hash value matches the fingerprint
   exchanged over SDP MUST be used.  If the certificate used in IKE does
   not match the original fingerprint, the endpoint MUST terminate the
   IKE session by detecting an authentication failure.

   In addition, each party MUST present a certificate and be
   authenticated by each other.

   The example described in Section 3 is for tunnel mode IPsec used for
   remote access, but the mode of negotiated IPsec is not limited to
   tunnel mode.  For example, IKE can negotiate transport mode IPsec to
   encrypt multiple media sessions between two parties with only a pair
   of IPsec security associations.  The only thing for which the SDP
   offer/answer model is responsible is to exchange the fingerprints of

certificates used for IKE; therefore, the SDP offer/answer is not
responsible for setting the security policy.

8. Specifications Assuming Prior Relationship between Two Nodes

   This section describes the specification for the limited cases in
   which certificates signed by trusted third parties or pre-shared keys
   between endpoints can be used for authentication in IKE.  Because the
   endpoints already have a prior relationship in this case, they use
   SIP servers for only name resolution and authorization.  However,
   even in this case, the integrity of the SDP description MUST be
   assured.

8.1. Certificates Signed by Trusted Third Party

   The protocol overview in this case is the same as in Section 3.  The
   SDP offer/answer procedure is also the same as in Sections 5 and 6.
   Both endpoints have a prior relationship through the trusted third
   parties, and SIP servers are used for name resolution and
   authorization of session initiation.  Even so, they MAY exchange
   fingerprints in the SDP because one device can have several
   certificates and it would be necessary to specify in advance which
   certificate will be used for the following IKE authentication.  This
   process also ensures that the certificate offered in the IKE process
   is the same as that owned by the peer that has been authorized at the
   SIP/SDP layer.  By this process, authorization in SIP and
   authentication in IKE become consistent with each other.

8.2. Configured Pre-Shared Key

   If a pre-shared key for IKE authentication is installed in both
   endpoints in advance, they need not exchange the fingerprints of
   their certificates.  However, they may still need to specify which
   pre-shared key they will use in the following IKE authentication in
   SDP because they may have several pre-shared keys.  Therefore, a new
   attribute, "psk-fingerprint", is defined to exchange the fingerprint
   of a pre-shared key over SDP.  This attribute also has the role of
   making authorization in SIP consistent with authentication in IKE.
   Attribute "psk-fingerprint" is applied to pre-shared keys as the
   "fingerprint" defined in [RFC4572] is applied to certificates.  The
   following is the ABNF of the "psk-fingerprint" attribute.  The use of
   "psk-fingerprint" is OPTIONAL.

   attribute                 =/ psk-fingerprint-attribute

   psk-fingerprint-attribute = "psk-fingerprint" ":" hash-func SP
                               psk-fingerprint

   hash-func                 = "sha-1" / "sha-224" / "sha-256" /
                               "sha-384" / "sha-512" / token
                               ; Additional hash functions can only come
                               ; from updates to RFC 3279

   psk-fingerprint           = 2UHEX *(":" 2UHEX)
                               ; Each byte in upper-case hex, separated
                               ; by colons.

   UHEX                      = DIGIT / %x41-46 ; A-F uppercase

   An example of SDP negotiation for IKE with pre-shared key
   authentication without IPsec NAT-Traversal is as follows.

   offer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.10
      a=ike-setup:active
      a=psk-fingerprint:SHA-1 \
      12:DF:3E:5D:49:6B:19:E5:7C:AB:4A:AD:B9:B1:3F:82:18:3B:54:02
      ...

   answer SDP
      ...
      m=application 500 udp ike-esp
      c=IN IP4 192.0.2.20
      a=ike-setup:passive
      a=psk-fingerprint:SHA-1 \
      12:DF:3E:5D:49:6B:19:E5:7C:AB:4A:AD:B9:B1:3F:82:18:3B:54:02
      ...

      Figure 6: SDP Example of IKE with Pre-Shared Key Authentication

9. Security Considerations

   This entire document concerns security, but the security
   considerations applicable to SDP in general are described in the SDP
   specification [RFC4566].  The security issues that should be
   considered in using comedia-tls are described in Section 7 in its
   specification [RFC4572].  This section mainly describes the security
   considerations specific to the negotiation of IKE using comedia-tls.

   Offering IKE in SDP (or agreeing to one in the SDP offer/answer
   model) does not create an obligation for an endpoint to accept any
   IKE session with the given fingerprint.  However, the endpoint must
   engage in the standard IKE negotiation procedure to ensure that the
   chosen IPsec security associations (including encryption and

   authentication algorithms) meet the security requirements of the
   higher-level application.  When IKE has finished negotiating, the
   decision to conclude IKE and establish an IPsec security association
   with the remote peer is entirely the decision of each endpoint.  This
   procedure is similar to how VPNs are typically established in the
   absence of SIP.

   In the general authentication process in IKE, subject DN or
   subjectAltName is recognized as the identity of the remote party.
   However, by using SIP identity and SIP-connected identity mechanisms
   in this spec, certificates are used simply as carriers for the public
   keys of the peers and there is no need for the information about who
   is the signer of the certificate and who is indicated by subject DN.

   In this document, the purpose of using IKE is to launch the IPsec SA;
   it is not for the security mechanism of RTP and RTCP [RFC3550]
   packets.  In fact, this mechanism cannot provide end-to-end security
   inside the VPN as long as the VPN uses tunnel mode IPsec.  Therefore,
   other security methods such as the Secure Real-time Transport
   Protocol (SRTP) [RFC3711] must be used to secure the packets.

   When using the specification defined in this document, it needs to be
   considered that under the following circumstances, security based on
   SIP authentication provided by SIP proxy may be breached.

   o  If a legitimate user's terminal is used by another person, it may
      be able to establish a VPN with the legitimate identity
      information.  This issue also applies to the general VPN cases
      based on the shared secret key.  Furthermore, in SIP we have a
      similar problem when file transfer, IM, or comedia-tls where non-
      voice/video is used as a means of communication.

   o  If a malicious user hijacks the proxy, he or she can use whatever
      credential is on the Access Control List (ACL) to gain access to
      the home network.

   For countermeasures to these issues, it is recommended to use unique
   information such as a password that only a legitimate user knows for
   VPN establishment.  Validating the originating user by voice or video
   before establishing VPN would be another method.

10. IANA Considerations

   IANA has registered the following new SDP attributes and media
   formats.

   Attribute name:         ike-setup
   Long form name:         IKE setup extensions
   Type of attribute:      Session-level and media-level
   Subject to charset:     No
   Purpose:                Attribute to indicate initiator and responder
                           of IKE-based media session
   Appropriate values:     See Section 4 of RFC 6193
   Contact name:           Makoto Saito, ma.saito@nttv6.jp

   Media format name:      ike-esp
   Long form name:         IKE followed by IPsec ESP
   Associated media:       application
   Associated proto:       udp
   Subject to charset:     No
   Purpose:                Media format that indicates IKE and IPsec ESP
                           as a VPN session
   Reference to the spec:  See Section 5 of RFC 6193
   Contact name:           Makoto Saito, ma.saito@nttv6.jp

   Media format name:      ike-esp-udpencap
   Long form name:         IKE followed by IPsec ESP or UDP encapsulated
                           IPsec ESP
   Associated media:       application
   Associated proto:       udp
   Subject to charset:     No
   Purpose:                Media format that indicates IKE that
                           supports NAT-Traversal and IPsec ESP or UDP
                           encapsulation of IPsec ESP packets as a VPN
                           session
   Reference to the spec:  See Section 5 of RFC 6193
   Contact name:           Makoto Saito, ma.saito@nttv6.jp

   Attribute name:         psk-fingerprint
   Long form name:         Fingerprint of pre-shared key extensions
   Type of attribute:      Session-level and media-level
   Subject to charset:     No
   Purpose:                Attribute to indicate a pre-shared key that
                           will be used in the following media session
   Appropriate values:     See Section 8.2. of RFC 6193
   Contact name:           Makoto Saito, ma.saito@nttv6.jp

11. Acknowledgments

   We would like to thank Remi Denis-Courmont, Dale Worley, Richard
   Barnes, David Hancock, Stuart Hoggan, Jean-Francois Mule, Gonzalo
   Camarillo, and Robert Sparks for providing comments and suggestions
   contributing to this document.  Eric Rescorla especially gave
   insightful comments from a security point of view.  Shintaro Mizuno
   and Shida Schubert also contributed a lot of effort to improving this
   document.

12. References

12.1. Normative References

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119, March 1997.

   [RFC3261]  Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston,
              A., Peterson, J., Sparks, R., Handley, M., and E.
              Schooler, "SIP: Session Initiation Protocol", RFC 3261,
              June 2002.

   [RFC3264]  Rosenberg, J. and H. Schulzrinne, "An Offer/Answer Model
              with Session Description Protocol (SDP)", RFC 3264, June
              2002.

   [RFC3947]  Kivinen, T., Swander, B., Huttunen, A., and V. Volpe,
              "Negotiation of NAT-Traversal in the IKE", RFC 3947,
              January 2005.

   [RFC3948]  Huttunen, A., Swander, B., Volpe, V., DiBurro, L., and M.
              Stenberg, "UDP Encapsulation of IPsec ESP Packets", RFC
              3948, January 2005.

   [RFC4301]  Kent, S. and K. Seo, "Security Architecture for the
              Internet Protocol", RFC 4301, December 2005.

   [RFC4303]  Kent, S., "IP Encapsulating Security Payload (ESP)", RFC
              4303, December 2005.

   [RFC4566]  Handley, M., Jacobson, V., and C. Perkins, "SDP: Session
              Description Protocol", RFC 4566, July 2006.

   [RFC4572]  Lennox, J., "Connection-Oriented Media Transport over the
              Transport Layer Security (TLS) Protocol in the Session
              Description Protocol (SDP)", RFC 4572, July 2006.

   [RFC5245]  Rosenberg, J., "Interactive Connectivity Establishment
              (ICE): A Protocol for Network Address Translator (NAT)
              Traversal for Offer/Answer Protocols", RFC 5245, April
              2010.

   [RFC5389]  Rosenberg, J., Mahy, R., Matthews, P., and D. Wing,
              "Session Traversal Utilities for NAT (STUN)", RFC 5389,
              October 2008.

   [RFC5996]  Kaufman, C., Hoffman, P., Nir, Y., and P. Eronen,
              "Internet Key Exchange Protocol Version 2 (IKEv2)", RFC
              5996, September 2010.

12.2. Informative References

   [RFC4474-Concerns]
              Rosenberg, J., "Concerns around the Applicability of RFC
              4474", Work in Progress, February 2008.

   [RFC3550]  Schulzrinne, H., Casner, S., Frederick, R., and V.
              Jacobson, "RTP: A Transport Protocol for Real-Time
              Applications", STD 64, RFC 3550, July 2003.

   [RFC3711]  Baugher, M., McGrew, D., Naslund, M., Carrara, E., and K.
              Norrman, "The Secure Real-time Transport Protocol (SRTP)",
              RFC 3711, March 2004.

   [RFC4028]  Donovan, S. and J. Rosenberg, "Session Timers in the
              Session Initiation Protocol (SIP)", RFC 4028, April 2005.

   [RFC4145]  Yon, D. and G. Camarillo, "TCP-Based Media Transport in
              the Session Description Protocol (SDP)", RFC 4145,
              September 2005.

   [RFC4474]  Peterson, J. and C. Jennings, "Enhancements for
              Authenticated Identity Management in the Session
              Initiation Protocol (SIP)", RFC 4474, August 2006.

   [RFC4916]  Elwell, J., "Connected Identity in the Session Initiation
              Protocol (SIP)", RFC 4916, June 2007.

   [RFC5246]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.2", RFC 5246, August 2008.

   [RFC5763]  Fischl, J., Tschofenig, H., and E. Rescorla, "Framework
              for Establishing a Secure Real-time Transport Protocol
              (SRTP) Security Context Using Datagram Transport Layer
              Security (DTLS)", RFC 5763, May 2010.

Authors' Addresses

   Makoto Saito
   NTT Communications
   1-1-6 Uchisaiwai-Cho, Chiyoda-ku
   Tokyo  100-8019
   Japan

   EMail: ma.saito@nttv6.jp

   Dan Wing
   Cisco Systems
   170 West Tasman Drive
   San Jose, CA  95134
   United States

   EMail: dwing@cisco.com

   Masashi Toyama
   NTT Corporation
   9-11 Midori-Cho 3-Chome, Musashino-Shi
   Tokyo  180-8585
   Japan

   EMail: toyama.masashi@lab.ntt.co.jp