SIP Working Group                                             Jari Arkko
INTERNET-DRAFT                                             Vesa Torvinen
<draft-ietf-sip-sec-agree-01.txt>                      Gonzalo Camarillo
May 2002                                                        Ericsson
Expires: December 2002                                        Tao Haukka
                                                                   Nokia
                                                              Sanjoy Sen
                                                         Nortel Networks


             Security Mechanism Agreement for SIP Sessions


Status of this memo

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

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

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

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

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

  This document is an individual submission to the IETF. Comments
  should be directed to the authors.


Abstract

  SIP has a number of security mechanisms. Some of them have been built
  in to the SIP protocol, such as HTTP authentication or secure
  attachments. These mechanisms have even alternative algorithms and
  parameters. SIP does not currently provide any mechanism for
  selecting which security mechanisms to use over a connection. In
  particular, even if some mechanisms such as OPTIONS were used to make
  this selection, the selection would be vulnerable against the
  Bidding-Down attack. This document defines three header fields for
  negotiating the security mechanisms within SIP between a user agent
  client and its next hop SIP entity. A SIP entity applying this
  mechanism must always require some minimum security (i.e. integrity
  protection) from all communicating parties in order to secure the
  negotiation, but the negotiation can agree on which specific minimum
  security is used.





Arkko et al                                                   [Page 1]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002



TABLE OF CONTENTS

  1.  Introduction....................................................2
  2.  The Problem.....................................................3
  3.  Solution........................................................4
      3.1. Requirements...............................................4
      3.2. Overview of Operations.....................................5
      3.3. Syntax.....................................................6
      3.4. Protocol Operation.........................................7
        3.4.1 Client Initiated........................................7
        3.4.2 Server Initiated........................................8
      3.5. Security Mechanism Initiation..............................9
      3.6. Duration of the Security Association......................10
      3.7. Summary of Header Field Use...............................10
  4.  Backwards Compatibility........................................11
  5.  Examples.......................................................11
      5.1. Client Initiated..........................................10
      5.2. Server Initiated..........................................12
  6.  Security Considerations........................................13
  7.  IANA Considerations............................................14
  8.  Modifications..................................................14
  9.  Acknowledgments................................................15
  10. Normative References...........................................15
  11. Non-Normative References.......................................15
  12. Authors’s Addresses............................................16


1. Introduction

  Traditionally, security protocols have included facilities to agree
  on the used mechanisms, algorithms, and other security parameters.
  The reason for this is that experience has shown that algorithm
  development uncovers problems in old algorithms and produces new
  ones. Furthermore, different mechanisms and algorithms are suitable
  for different situations. Typically, protocols also select other
  parameters beyond algorithms at the same time.

  The purpose of this specification is to define a similar negotiation
  functionality in SIP [1]. SIP has some security functionality built-
  in such as HTTP Digest authentication [4], secure attachments such as
  S/MIME [5], and can also use underlying security protocols such as
  IPsec/IKE [2] or TLS [3]. Some of the built-in security functionality
  allows also alternative algorithms and other parameters. While some
  work within the SIP Working Group has been looking towards reducing
  the number of recommended security solutions (i.e., recommend just
  one lower layer security protocol), we can not expect to cut down the
  number of items in the whole list to one. There will still be
  multiple security solutions utilized by SIP. Furthermore, it is
  likely that new  methods will appear in the future, to complement the
  methods that exist today.

  Chapter 2 shows that without a secured method to choose between
  security mechanisms and/or their parameters, SIP is vulnerable to
  certain attacks. As the HTTP authentication RFC [4] points out,
  authentication and integrity protection using multiple alternative



Arkko et al                                                   [Page 2]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  methods and algorithms is vulnerable to Man-in-the-Middle (MitM)
  attacks. More seriously, it is hard or sometimes even impossible to
  know whether a SIP peer entity is truly unable to perform (e.g.,
  Digest, TLS, or S/MIME) or if a MitM attack is in action. In small
  networks consisting of workstations and servers these issues are not
  very relevant, as the administrators can deploy appropriate software
  versions and set up policies for using exactly the right type of
  security. However, SIP will be deployed to hundreds of millions of
  small devices with little or no possibilities for coordinated
  security policies, let alone software upgrades, and this makes these
  issues much worse. This conclusion is also supported by the
  requirements from 3GPP [6].

  Chapter 6 documents the proposed solution, and chapter 7 gives some
  demonstrative examples.

2. Problem Description

  SIP has alternative security mechanisms such as HTTP authentication
  with integrity protection, lower layer security protocols, and
  S/MIME. It is likely that their use will continue in the future. SIP
  security is developing, and is likely to see also new solutions in
  the future.

  Deployment of large number of SIP-based consumer devices such as 3GPP
  terminals requires all network devices to be able to accommodate
  past, current and future mechanisms; there is no possibility for
  instantaneous change since the new solutions are coming gradually in
  as new standards and product releases occur. It is sometimes even
  impossible to upgrade some of the devices without getting completely
  new hardware.

  So, the basic security problem that such a large SIP-based network
  must consider, would be on how do security mechanisms get selected?
  It would be desirable to take advantage of new mechanisms as they
  become available in products.

  Firstly, we need to know somehow what security should be applied, and
  preferably find this out without too many additional roundtrips.

  Secondly, selection of security mechanisms MUST be secure.
  Traditionally, all security protocols use a secure form of
  negotiation. For instance, after establishing mutual keys through
  Diffie-Hellman, IKE sends hashes of the previously sent data --
  including the offered crypto mechanisms. This allows the peers to
  detect if the initial, unprotected offers were tampered with.

  The security implications of this are subtle, but do have a
  fundamental importance in building large networks that change over
  time. Given that the hashes are produced also using algorithms agreed
  in the first unprotected messages, one could ask what the difference
  in security really is. Assuming integrity protection is mandatory and
  only secure algorithms are used, we still need to prevent MitM
  attackers from modifying other parameters, such as whether encryption
  is provided or not. Let us first assume two peers capable of using
  both strong and weak security. If the initial offers are not



Arkko et al                                                   [Page 3]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  protected in any way, any attacker can easily "downgrade" the offers
  by removing the strong options. This would force the two peers to use
  weak security between them. But if the offers are protected in some
  way -- such as by hashing, or repeating them later when the selected
  security is really on -- the situation is different. It would not be
  sufficient for the attacker to modify a single message. Instead, the
  attacker would have to modify both the offer message, as well as the
  message that contains the hash/repetition. More importantly, the
  attacker would have to forge the weak security that is present in the
  second message, and would have to do so in real time between the sent
  offers and the later messages. Otherwise, the peers would notice that
  the hash is incorrect. If the attacker is able to break the weak
  security, the security method and/or the algorithm should not be
  used.

  In conclusion, the security difference is making a trivial attack
  possible versus demanding the attacker to break algorithms. An
  example of where this has a serious consequence is when a network is
  first deployed with integrity protection (such as HTTP Digest [4]),
  and then later new devices are added that support also encryption
  (such as S/MIME [1]). In this situation, an insecure negotiation
  procedure allows attackers to trivially force even new devices to use
  only integrity protection.

3. Solution

3.1 Requirements

  The solution to the SIP security negotiation problem should have the
  following properties:

  (a) It allows the selection of security mechanisms, such as lower
  layer security protocols or HTTP digest. It also allows the selection
  of individual algorithms and parameters when the security functions
  are integrated in SIP (such as in the case of HTTP authentication).

  (b) It allows first-hop security negotiation.

  (c) It is secure (i.e., prevents the bidding down attack.)

  (d) It is capable of running without additional roundtrips. This is
  important in the cellular environment, where an additional roundtrip
  could delay the call set up for 1000-1500 ms.

  (e) It does not introduce any additional state to servers and
  proxies.

  Currently, SIP does not have any mechanism which fulfills all the
  requirements above. The basic SIP features such as OPTIONS and
  Require, Supported headers are capable of informing peers about
  various capabilities including security mechanisms. However, the
  straight forward use of these features can not guarantee a secured
  agreement. HTTP Digest algorithm lists [4] are not secure for picking
  among the digest integrity algorithms, as is described in the RFC
  itself. More seriously, they have no provisions for allowing




Arkko et al                                                   [Page 4]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  encryption to be negotiated. Hence, it would be hard to turn on
  possible future encryption schemes in a secure manner.

  A self-describing security mechanism is a security mechanism that,
  when used, contains all necessary information about the method being
  used as well as all of its parameters such as algorithms.

  A non-self-describing security mechanism is a security mechanism
  that, when used, requires that the use of the method or some of its
  parameters have been agreed beforehand.

  Most security mechanisms used with SIP are self-describing. The use
  of HTTP digest, as well as the chosen algorithm is visible from the
  HTTP authentication headers. The use of S/MIME is indicated by the
  MIME headers, and the CMS structures inside S/MIME describe the used
  algorithms. TLS is run on a separate port in SIP, and where IPsec/IKE
  is used, IKE negotiates all the necessary parameters.

  The only exception to this list is the use of manually keyed IPsec.
  IPsec headers do not contain information about the used algorithms.
  Furthermore, peers have to set up IPsec Security Associations before
  they can be used to receive traffic. In contrast S/MIME can be
  received even if no Security Association was in place, because the
  application can search for a Security Association (or create a new
  one) after having received a message that contains S/MIME.

  In order to make it possible to negotiate both self-describing and
  non-self-describing security mechanisms, we need another requirement
  on the security agreement scheme:

  (f) the security agreement scheme must allow both sides to decide on
  the desired security mechanism before it is actually used.

  This decision can, and must, take place on both sides before we can
  be sure that the negotiation has not been tampered by a man-in-the-
  middle. This tampering will be detected later.

3.2. Overview of Operations

  The message flow below illustrates how the mechanism defined in this
  document works:

         1. Client ----------client list---------> Server
         2. Client <---------server list---------- Server
         3. Client ------(turn on security)------- Server
         4. Client ----------server list---------> Server
         5. Client <---------ok or error---------- Server

       Figure 1: Security negotiation message flow

  Step 1: Clients wishing to use this specification can send a list of
  their supported security mechanisms along the first request to the
  server.

  Step 2: Servers wishing to use this specification can challenge the
  client to perform the security agreement procedure. The security



Arkko et al                                                   [Page 5]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  mechanisms and parameters supported by the server are sent along in
  this challenge.

  Step 3: The client then proceeds to select the highest-preference
  security mechanism they have in common and to turn on the selected
  security.

  Step 4: The client contacts the server again, now using the selected
  security mechanism. The server's list of supported security
  mechanisms is returned as a response to the challenge.

  Step 5: The server verifies its own list of security mechanisms in
  order to ensure that the original list had not been modified.

  This procedure is stateless for servers (unless the used security
  mechanisms require the server to keep some state).

  The client and the server lists are both static (i.e., they do not
  and cannot change based on the input from the other side). Nodes may,
  however, maintain several static lists, one for each interface, for
  example.

  Between Steps 1 and 2, the server may set up a non-self-describing
  security mechanism if necessary. Note that with this type of security
  mechanisms, the server is necessarily stateful. The client would set
  up the non-self-describing security mechanism between Steps 2 and 4.

3.3. Syntax

  We define three new SIP header fields, namely Security-Client,
  Security-Server and Security-Verify. Their BNF syntax is provided
  below:

     security-client = "Security-Client" HCOLON
                       sec-mechanism *(COMMA sec-mechanism)
     security-server = "Security-Server" HCOLON
                       sec-mechanism *(COMMA sec-mechanism)
     security-verify = "Security-Verify" HCOLON
                       sec-mechanism *(COMMA sec-mechanism)
     sec-mechanism   = mechanism-name *(SEMI mech-parameters)
     mechanism-name  = ( "digest-integrity" / "tls" / "ipsec-ike" /
                        "ipsec-man" / "smime" / token )
     mech-parameters = ( preference / algorithm / extension )
     preference      = “q” EQUAL qvalue
     qvalue          = ( “0” [ “.” 0*3DIGIT ] )
                        / ( “1” [ “.” 0*3(“0”) ] )
     algorithm       = "alg" EQUAL token
     extension       = generic-param

  Note that qvalue is already defined in the SIP BNF [1]. We have
  copied its definitions here for completeness.

  The parameters described by the BNF above have the following
  semantics:





Arkko et al                                                   [Page 6]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


    Mechanism-name: It identifies the security mechanism supported by
    the client, when it appears in a security-client header fields, or
    by the server, when it appears in a security-server header field.
    This specification defines six values:

      - "tls" for TLS [3].
      - "digest-integrity" for HTTP Digest [4] using additional
      integrity protection (i.e., the qop parameter) for the Security-
      Verify header field.
      - "ipsec-ike" for IPsec with IKE [2].
      - "ipsec-man" for manually keyed IPsec without IKE.
      - "smime" for S/MIME [5].


    Preference: The "q" value indicates a relative preference for the
    particular mechanism. The higher the value the more preferred the
    mechanism is.

    Algorithm: An optional algorithm field for those security
    mechanisms which are not self-describing or which are vulnerable
    for bidding-down attacks (e.g., HTTP Digest). In the case of HTTP
    Digest, the same rules apply as defined in [4] for the "algorithm"
    field in HTTP Digest.

3.4. Protocol Operation

  This section deals with the protocol details involved in the
  negotiation between a user agent client and its next-hop SIP entity.
  Throughout the text the next-hop SIP entity is referred to as the
  first-hop proxy or outbound proxy. However, the reader should bear in
  mind that a user agent server can also be the next-hop for a user
  agent client in the absence of proxies. Note as well that a proxy can
  also have an outbound proxy.

3.4.1 Client Initiated

  A client wishing to establish some type of security with its first-
  hop proxy SHOULD add a Security-Client header field to a request
  addressed to this proxy (i.e., the destination of the request is the
  first-hop proxy). This header field contains a list of all the
  security mechanisms that the client supports. The client SHOULD NOT
  add preference parameters to this list. The client MUST also add a
  Require header field with the value "sec-agree" to its request.

  The Security-Client header field is used by the server to include any
  necessary information in its response. For example, if digest-
  integrity is the chosen mechanism, the server includes a WWW-
  Authenticate header in the response. If S/MIME is chosen, the
  appropriate certificate is included. If the security mechanisms
  supported by the client do not need any further information to be
  established (e.g., TLS) the client MAY choose not to include the
  Security-Client header field in its request.

  A server receiving a request that contains a Require header field
  with the value "sec-agree" MUST challenge the client with a 494
  (Security Agreement Required) response. The server MUST add a



Arkko et al                                                   [Page 7]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  Security-Server header field to this response listing the security
  mechanisms that the server supports. The server MUST add its list to
  the response even if there are no common security mechanisms in the
  client's and server's lists. The server’s list MUST NOT depend on the
  contents of the client's list.

  The server MUST compare the list received in the Security-Client
  header field with the list to be sent in the Security-Server header
  field. When the client receives this response, it will choose the
  common security mechanism with the higher preference value.
  Therefore, the server MUST add the necessary information so that the
  client can initiate that mechanism (e.g., a WWW-Authenticate header
  field for digest-integrity).

  When the client receives a response with a Security-Server header
  field, it SHOULD choose the security mechanism in the server’s list
  with the highest "q" value among all the mechanisms that are known to
  the client. Then, it MUST initiate that particular security mechanism
  as described in Section 3.5. This initiation may be carried out
  without involving any SIP message exchange (e.g., establishing a TLS
  connection).

  All the subsequent SIP requests sent by the client SHOULD make use of
  the security mechanism initiated in the previous step. These requests
  MUST contain a Security-Verify header field that mirrors the server’s
  list received previously in the Security-Server header field. This
  request MAY use SIP loose routing mechanism (i.e., Route header
  fields) to traverse the proxy, but its final destination may be
  different than the proxy. In this case, the request SHOULD NOT
  include a Require header field with the value "sec-agree".

    For example, the first request was an OPTIONS request directly
    addressed to the proxy and the second request is an INVITE that
    will traverse the proxy but that is addressed to a real user (see
    example in section 4.1).

  The server MUST check that the security mechanisms listed in the
  Security-Verify header field of incoming requests correspond to its
  static list of supported security mechanisms.  The server can proceed
  processing a particular request if, and only if, the list was not
  modified.  If modification of the list is detected, the server MUST
  challenge the client with a 494 (Security Agreement Required). This
  response MUST include a challenge with server's unmodified list of
  supported security mechanisms.

  Once the security has been negotiated between two SIP entities, the
  same SIP entities MAY use the same security when communicating with
  each other in different SIP roles. For example, if a UAC and its
  outbound proxy negotiate some security, they may try to use the same
  security for incoming requests (i.e., the UA will be acting as a
  UAS).

  The user of a UA MAY be informed about the results of the security
  mechanism negotiation. The user MAY decline to accept a particular
  security mechanism, and abort further SIP communications with the
  peer.



Arkko et al                                                   [Page 8]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002



3.4.2 Server Initiated

  A server decides to use the security negotiation described in this
  document based on local policy. A server that decides to use this
  negotiation MUST challenge requests regardless of the presence or the
  absence of any Require or Supported header fields in incoming
  requests.

  A server that by policy requires the use of this specification and
  receives a request that does not have the sec-agree option tag in a
  Require or Supported header field MUST return a 421 (Extension
  Required) response. If the request had the sec-agree option tag in a
  Supported header field, it MUST return a 494 (Security Agreement
  Required) response. In both situation the server MUST also include in
  the response a Security-Server header field listing its capabilities
  and a Require header field with an option-tag 'sec-agree' in it. All
  the Via header field entries in the response except the topmost value
  MUST be removed.

  Clients that support the extension defined in this document MAY add a
  Supported header field with a value of "sec-agree". In addition to
  this, clients SHOULD add a Security-Client header field so that they
  can save a round trip in case the server decides to challenge the
  request.

3.5. Security mechanism initiation

  Once the client chooses a security mechanism from the list received
  in the Security-Server header field from the server, it initiates
  that mechanism. Different mechanisms require different initiation
  procedures.

  If TLS is chosen, the client MUST contact the server using the host
  part of the Request-URI in the first request to the server as the
  destination of the connection (note that this may involve using
  standard SIP DNS procedures to locate a server). If this connection
  attempt fails, the security agreement procedure MUST be considered to
  have failed, and MUST be terminated.

  If digest-integrity is chosen, the 494 (Security Agreement Required)
  response will contain an HTTP authentication challenge. The client
  MUST use the qos parameter possibly together with some variant of
  MIME tunneling so that the Security-Verify header field in the
  request is integrity protected in the MIME body. Note that digest
  alone would not fulfill the minimum security requirements of this
  specification.

  To use "ipsec-ike", the client attempts to establish an IKE
  connection to the host part of the Request-URI in the first request
  to the server. If the IKE connection attempt fails, the agreement
  procedure MUST be considered to have failed, and MUST be terminated.

  Note that "ipsec-man" will only work if the communicating SIP
  entities know which keys and other parameters to use. It is outside




Arkko et al                                                   [Page 9]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  the scope of this specification to describe how this information can
  be made known to the peers.

  In both IPsec-based mechanisms, it is expected that appropriate
  policy entries for protecting SIP have been configured or will be
  created before attempting to use the security agreement procedure,
  and that SIP communications use port numbers and addresses according
  to these policy entries. It is outside the scope of this
  specification to describe how this information can be made known to
  the peers, but it could be typically configured at the same time as
  the IKE credentials or manual SAs have been entered.

  To use S/MIME, the client MUST construct its request using S/MIME.
  The client may have received the server’s certificate in an S/MIME
  body in the 494 (Security Agreement Required) response.

3.6. Duration of Security Associations

  Once a security mechanism has been negotiated, both the server and
  the client need to know until when it can be used. All the mechanisms
  described in this document have a different way to signal the end of
  a security association. When TLS is used, the termination of the
  connection indicates that a new negotiation is needed. IKE negotiates
  the duration of a security association. If the credentials provided
  by a client using digest-integrity are not longer valid, the server
  will re-challenge the client. It is assumed that when IPsec-man is
  used, the same out-of-band mechanism used to distribute keys is used
  to define the duration of the security association.

3.7. Summary of Header Field Use

  The header fields defined in this document may be used to negotiate
  the security mechanisms between a UAC and other SIP entities
  including UAS, proxy, and registrar. Information about the use of
  headers in relation to SIP methods and proxy processing is summarized
  in Table 1.

  Header field           where        proxy ACK BYE CAN INV OPT REG

  _________________________________________________________________
  Security-Client          R           ard   -   o   -   o   o   o
  Security-Server     401,407,421,494        -   o   -   o   o   o
  Security-Verify          R           ard   -   o   -   o   o   o


  Header field           where        proxy SUB NOT PRK IFO UPD MSG

  _________________________________________________________________
  Security-Client          R           ard   o   o   -   o   o   o
  Security-Server     401,407,421,494        o   o   -   o   o   o
  Security-Verify          R           ard   o   o   -   o   o   o

                    Table 1: Summary of header usage.


  The "where" column describes the request and response types in which



Arkko et al                                                  [Page 10]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  the header field may be used. The header may not appear in other
  types of SIP messages. Values in the where column are:

  - R: Header field may appear in requests.

  - 401, 407 etc.: A numerical value or range indicates response codes
    with which the header field can be used.

  The "proxy" column describes the operations a proxy may perform on a
  header field:

  - a: A proxy can add or concatenate the header field if not present.

  - r: A proxy must be able to read the header field, and thus this
       header field cannot be encrypted.

  - d: A proxy can delete a header field value.

  The next six columns relate to the presence of a header field in a
  method:


  - o: The header field is optional.

4. Backwards Compatibility

  A server that, by local policy, decides to use the negotiation
  mechanism defined in this document, will not accept requests from
  clients that do not support this extension. This obviously breaks
  interoperability with every plain SIP client. Therefore, this
  extension should only be used in closed environments where it is
  ensured somehow that every client implements this extension.

5. Examples

  The following examples illustrate the use of the mechanism defined
  above.

5.1. Client Initiated

  A UA negotiates the security mechanism to be used with its outbound
  proxy without knowing beforehand which mechanisms the proxy supports.


          UAC                 Proxy               UAS

           |                    |                  |
           |----(1) OPTIONS---->|                  |
           |                    |                  |
           |<-----(2) 494-------|                  |
           |                    |                  |
           |<=======TLS========>|                  |
           |                    |                  |
           |----(3) INVITE----->|                  |
           |                    |----(4) INVITE--->|
           |                    |                  |



Arkko et al                                                  [Page 11]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


           |                    |<---(5) 200 OK----|
           |<---(6) 200 OK------|                  |
           |                    |                  |
           |------(7) ACK------>|                  |
           |                    |-----(8) ACK----->|
           |                    |                  |
           |                    |                  |
           |                    |                  |
           |                    |                  |

         Figure 2: Negotiation initiated by the client

  The UAC sends an OPTIONS request to its outbound proxy, indicating
  that it is able to negotiate security mechanisms and that it supports
  TLS and digest-integrity (Step 1 of figure 1). The outbound proxy
  challenges the UAC with its own list of security mechanisms – IPsec
  and TLS (Step 2 of figure 1). The only common security mechanism is
  TLS, so they establish a TLS connection between them (Step 3 of
  figure 1). When the connection is successfully established, the UAC
  sends an INVITE over the TLS connection just established (Step 4 of
  figure 1). This INVITE contains the server’s security list. The
  server verifies it, and since it matches its static list, it
  processes the INVITE and forwards it to the next hop.

  If this example was run without Security-Server header in Step 2, the
  UAC would not know what kind of security the other one supports, and
  would be forced to error-prone trials.

  More seriously, if the Security-verify was omitted in Step 4, the
  whole process would be prone for MitM attacks. An attacker could
  spoof "ICMP Port Unreachable" message on the trials, or remove the
  stronger security option from the header in Step 1, therefore
  substantially reducing the security.

        (1) OPTIONS proxy.example.com
            Security-Client: tls;q=0.1
            Security-Client: digest-integrity;q=0.2
            Require: sec-agree

        (2) 494 (Security Agreement Required)
            Security-Server: ipsec-ike;q=0.1
            Security-Server: tls;q=0.2

        (3) INVITE proxy.example.com
            Security-Verify: ipsec-ike;q=0.1
            Security-Verify: tls;q=0.2
            Route: callee@domain.com

  The 200 OK response for the INVITE and the ACK are also sent over the
  TLS connection. The ACK (7) will contain the same Security-Verify
  header field as the INVITE (3).

5.2. Server Initiated






Arkko et al                                                  [Page 12]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  In this example of figure 3 the client sends an INVITE towards the
  callee using an outbound proxy. This INVITE does not contain any
  Require header field.


         UAC                 Proxy               UAS

          |                    |                  |
          |-----(1) INVITE---->|                  |
          |                    |                  |
          |<-----(2) 421-------|                  |
          |                    |                  |
          |------(3) ACK------>|                  |
          |                    |                  |
          |<=======IKE========>|                  |
          |                    |                  |
          |-----(4) INVITE---->|                  |
          |                    |----(5) INVITE--->|
          |                    |                  |
          |                    |<---(7) 200 OK----|
          |<----(6) 200 OK-----|                  |
          |                    |                  |
          |------(8) ACK------>|                  |
          |                    |-----(9) ACK----->|
          |                    |                  |
          |                    |                  |

       Figure 3: Server initiated security negotiation

  The proxy, following its local policy, challenges the INVITE. It
  returns a 421 (Extension Required) with a Security-Server header
  field that lists IPsec-IKE and TLS. Since the UAC supports IPsec-IKE
  it performs the key exchange and establishes a security association
  with the proxy. The second INVITE (4) and the ACK (8) contain a
  Security-Verify header field that mirrors the Security-Server header
  field received in the 421. The INVITE (4), the 200 OK (6) and the ACK
  (8) are sent using the security association that has been
  established.


6. Security Considerations

  This specification is about making it possible to select between
  various SIP security mechanisms in a secure manner. In particular,
  the method presented here allow current networks using, for instance,
  Digest, to be securely upgraded to, for instance, IPsec without
  requiring a simultaneous modification in all equipment. The method
  presented in this specification is secure only if the weakest
  proposed mechanism offers at least integrity protection.

  Attackers could try to modify the server's list of security
  mechanisms in the first response. This would be revealed to the
  server when the client returns the received list using the security.

  Attackers could also try to modify the repeated list in the second
  request from the client. However, if the selected security mechanism



Arkko et al                                                  [Page 13]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  uses encryption this may not be possible, and if it uses integrity
  protection any modifications will be detected by the server.

  Finally, attackers could try to modify the client's list of security
  mechanisms in the first message. The client selects the security
  mechanism based on its own knowledge of its own capabilities and the
  server's list, hence the client's choice would be unaffected by any
  such modification. However, the server's choice could still be
  affected as described below:

  - If the modification affected the server's choice, the server and
  client would end up choosing different security mechanisms in Step 3
  or 4 of figure 1. Since they would be unable to communicate to each
  other, this would be detected as a potential attack. The client would
  either retry or give up in this situation.

  - If the modification did not affect the server's choice, there's no
  effect.

  All clients that implement this specification MUST select HTTP Digest
  with integrity, TLS, IPsec, or any stronger method for the protection
  of the second request. If HTTP Digest is used alone, the security
  agreement headers MUST be protected. This can be done with HTTP
  Digest if combined with MIME/SIP tunneling, for example.

7. IANA Considerations

  This specification defines the 'sec-agree' SIP option tag which
  should be registered in IANA.

  This specification also defines a new SIP status code, 494 (Security
  Agreement Failed), which should be registered in IANA.

8. Modifications

  The draft-sip-sec-agree-01.txt version of this specification
  introduced the following modifications:

   - Scope narrowed down to first-hop negotiation.

   - Fixed syntax of header fields.

  The draft-sip-sec-agree-00.txt version of this specification
  introduced the following modifications:

    - Many editorial changes, restructuring and clarifications.

    - Motivation section has been shortened since this is now a WG
    item.

    - Clarified that the solution requires always some base level of
    security (i.e. integrity) in order to work. Even 'the weak
    security' must not be broken.

    - Text related to alternative solutions shortened and moved to a
    new place.



Arkko et al                                                  [Page 14]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002



    - New rules for possible error and special cases has been added,
    (e.g., for the case in which an non-adjacent SIP entities try to
    negotiate hop-by-hop security mechanisms).

    - Syntax of the header redesigned. Wanted to get rid of the
    semantics related to the relative position of a header component in
    the header e.g., first parameters defines the 'from-uri', second
    the 'to-uri', and third the first supported security mechanism).
    The option tags are now used to identify the Security Agreement
    extension, not the individual security mechanisms.

    - The semantics of the header slightly changed: the AND operator
    between the indivicual mechanisms is removed because it is really
    need with HTTP Digest only. And even in this case, the negotiation
    is not needed beforehand if some underlying security is used.

    - Options for HTTP Digest algorithms and manually keyed IPsec
    added.

    - Explicit rules were added to all mechanisms on how they should be
    used, such as TLS to be run on port 5061 etc.

    - References to Enhanced HTTP Digest removed.

    - Example related to 3GPP generalized.

  The draft-arkko-sip-sec-agree-01.txt version of this specification
  introduced the following modifications:

    - Reversed approach to make servers stateless

    - Removed discussion of the use of this for Digest algorithm
    selection, since Enhanced Digest already has bidding-down
    protection

    - Renamed org.iana.sip.digest to org.iana.sip.edigest and removed
    the parameters, as we can rely on Enhanced Digest to perform the
    algorithm selection.

    - Removed agreements for full paths.

    - Simplified syntax

9. Acknowledgments

  The authors wish to thank Lee Valerius, Rolf Blom, James Undery,
  Jonathan Rosenberg, Hugh Shieh, Gunther Horn, Krister Boman, David
  Castellanos-Zamora, Aki Niemi, Miguel Garcia, Valtteri Niemi, Martin
  Euchner, Eric Rescorla and members of the 3GPP SA3 group for
  interesting discussions in this problem space.



10. Normative References




Arkko et al                                                  [Page 15]


INTERNET-DRAFT             SIP Sec Agreement                   May 2002


  [1] J. Rosenberg, H. Schulzrinne, G. Camarillo, A. Johnston, J.
  Peterson, R. Sparks, M. Handley, E. Schooler "SIP: Session Initiation
  Protocol", Work In Progress, draft-ietf-sip-rfc2543bis-09.txt, IETF,
  February 2002.

  [2] S. Kent, R. Atkinson, "Security Architecture for the Internet
  Protocol", RFC 2401, IETF, November 1998.

  [3] T. Dierks, C. Allen, "The TLS Protocol Version 1.0", RFC 2246,
  IETF January 1999.

  [4] Franks, J. et al, "HTTP Authentication: Basic and Digest Access
  Authentication", RFC 2617, IETF, June 1999.

  [5] B. Ramsdell and Ed, "S/MIME version 3 message specification," RFC
  2633, IETF, June 1999.

11. Non-Normative References

  [6] M. Garcia, D. Mills, G. Bajko, G. Mayer, F. Derome, H. Shieh, A.
  Allen, S. Chotai, K. Drage, J. Bharatia, "3GPP requirements on SIP",
  draft-garcia-sipping-3gpp-reqs-00.txt. Work In Progress, IETF,
  October 2001.

12.  Authors's Addresses

  Jari Arkko
  Ericsson
  02420 Jorvas
  Finland
  EMail: Jari.Arkko@ericsson.com

  Vesa Torvinen
  Ericsson
  02420 Jorvas
  Finland
  EMail: Vesa.Torvinen@ericsson.fi

  Gonzalo Camarillo
  Ericsson
  02420 Jorvas
  Finland
  EMail: Gonzalo.Camarillo@ericsson.com

  Tao Haukka
  Nokia
  Finland
  EMail: Tao.Haukka@nokia.com

  Sanjoy Sen
  Nortel Networks
  2735-B Glenville Drive
  Richardson, TX 75082, USA
  EMail: sanjoy@nortelnetworks.com





Arkko et al                                                  [Page 16]