One document matched: draft-ietf-avtcore-srtp-ekt-01.xml


<?xml version="1.0" encoding="US-ASCII"?>
<!DOCTYPE rfc SYSTEM "rfc2629.dtd" [
<!ENTITY rfc2119 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2119.xml">
<!ENTITY rfc5226 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5226.xml">
<!ENTITY rfc6347 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6347.xml">
<!ENTITY rfc6407 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.6407.xml">
<!ENTITY rfc5234 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5234.xml">
<!ENTITY rfc2675 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2675.xml">
<!ENTITY rfc4301 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4301.xml">
<!ENTITY rfc2409 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2409.xml">
<!ENTITY rfc2410 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2410.xml">
<!ENTITY rfc2406 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2406.xml">
<!ENTITY rfc2407 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.2407.xml">
<!ENTITY rfc3261 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3261.xml">
<!ENTITY rfc3264 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3264.xml">
<!ENTITY rfc5649 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5649.xml">
<!ENTITY rfc4648 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4648.xml">
<!ENTITY rfc3550 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3550.xml">
<!ENTITY rfc3610 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3610.xml">
<!ENTITY rfc3602 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3602.xml">
<!ENTITY rfc3686 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3686.xml">
<!ENTITY rfc3830 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3830.xml">
<!ENTITY rfc3711 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.3711.xml">
<!ENTITY rfc4563 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4563.xml">
<!ENTITY rfc4771 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4771.xml">
<!ENTITY rfc4567 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4567.xml">
<!ENTITY rfc4568 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4568.xml">
<!ENTITY rfc5764 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.5764.xml">
<!ENTITY rfc4566 PUBLIC "" "http://xml.resource.org/public/rfc/bibxml/reference.RFC.4566.xml">
]>
<?xml-stylesheet type='text/xsl' href='rfc2629.xslt' ?>
<?rfc toc="yes" ?>
<?rfc symrefs="yes" ?>
<?rfc sortrefs="yes"?>
<?rfc iprnotified="no" ?>
<?rfc strict="yes" ?>
<rfc category="std" docName="draft-ietf-avtcore-srtp-ekt-01" ipr="trust200902">
  <front>
    <title abbrev="EKT SRTP">Encrypted Key Transport for Secure RTP</title>

    <author fullname="David A. McGrew" initials="D.A.M." surname="McGrew">
      <organization abbrev="Cisco">Cisco Systems, Inc.</organization>

      <address>
        <postal>
          <street>510 McCarthy Blvd.</street>

          <city>Milpitas</city>

          <region>CA</region>

          <code>95035</code>

          <country>US</country>
        </postal>

        <phone>(408) 525 8651</phone>

        <email>mcgrew@cisco.com</email>

        <uri>http://www.mindspring.com/~dmcgrew/dam.htm</uri>
      </address>
    </author>

    <author fullname="Flemming Andreason" initials="F.A." surname="Andreasen">
      <organization abbrev="Cisco">Cisco Systems, Inc.</organization>

      <address>
        <postal>
          <street>499 Thornall Street</street>

          <city>Edison</city>

          <region>NJ</region>

          <code>08837</code>

          <country>US</country>
        </postal>

        <email>fandreas@cisco.com</email>
      </address>
    </author>

    <author fullname="Dan Wing" initials="D." surname="Wing">
      <organization abbrev="Cisco">Cisco Systems, Inc.</organization>

      <address>
        <postal>
          <street>510 McCarthy Blvd.</street>

          <city>Milpitas</city>

          <region>CA</region>

          <code>95035</code>

          <country>US</country>
        </postal>

        <phone>(408) 853 4197</phone>

        <email>dwing@cisco.com</email>
      </address>
    </author>

<!--
    <author fullname="Kai Fischer" initials="K." surname="Fischer">
      <organization abbrev="Siemens Enterprise Communications">Siemens
      Enterprise Communications GmbH & Co. KG</organization>

      <address>
        <postal>
          <street>Hofmannstr. 51</street>

          <city>Munich</city>

          <region>Bavaria</region>

          <code>81739</code>

          <country>Germany</country>
        </postal>

        <email>kai.fischer@siemens-enterprise.com</email>
      </address>
    </author>
-->

    <date />

    <area>RAI</area>

    <workgroup>AVTCORE Working Group</workgroup>

    <keyword>RTP</keyword>

    <keyword>SRTP</keyword>

    <keyword>EKT</keyword>

    <abstract>
      <t>Encrypted Key Transport (EKT) is an extension to Secure
      Real-time Transport Protocol (SRTP) that provides for the secure
      transport of SRTP master keys, Rollover Counters, and other
      information, within SRTP or SRTCP. This facility enables SRTP to
      work for decentralized conferences with minimal control.
      </t>

      <t>This note defines EKT, and also describes how to use it with
      SDP Security Descriptions, DTLS-SRTP, and MIKEY. These other key
      management protocols provide an EKT key to everyone in a
      session, and EKT coordinates the keys within the session.</t>
    </abstract>
  </front>

  <middle>
    <section title="Introduction">
      <t>RTP is designed to allow decentralized groups with minimal
      control to establish sessions, such as for multimedia
      conferences.   Unfortunately, Secure RTP (<xref
      target="RFC3711">SRTP</xref>) cannot be used in many
      minimal-control scenarios, because it requires that SSRC values
      and other data be coordinated among all of the participants in a
      session.  For example, if a participant joins a session that is
      already in progress, the SRTP rollover counter (ROC) of each
      SRTP source in the session needs to be provided to that
      participant.</t>

      <t>The inability of SRTP to work in the absence of central
      control was well understood during the design of that protocol;
      that omission was considered less important than optimizations
      such as bandwidth conservation. Additionally, in many situations
      SRTP is used in conjunction with a signaling system that can
      provide most of the central control needed by SRTP. However,
      there are several cases in which conventional signaling systems
      cannot easily provide all of the coordination required. It is
      also desirable to eliminate the layer violations that occur when
      signaling systems coordinate certain SRTP parameters, such as
      SSRC values and ROCs.</t>

      <t>This document defines Encrypted Key Transport (EKT) for SRTP, an
      extension to SRTP that fits within the SRTP framework and reduces the
      amount of signaling control that is needed in an SRTP session. EKT
      securely distributes the SRTP master key and other information for each
      SRTP source, using SRTCP or SRTP to transport that information. With
      this method, SRTP entities are free to choose SSRC values as they see
      fit, and to start up new SRTP sources with new SRTP master keys (see
      Section 2.2) within a session without coordinating with other entities
      via signaling or other external means. This fact allows to reinstate the
      RTP collision detection and repair mechanism, which is nullified by the
      current SRTP specification because of the need to control SSRC values
      closely. An SRTP endpoint using EKT can generate new keys whenever an
      existing SRTP master key has been overused, or start up a new SRTP
      source to replace an old SRTP source that has reached the packet-count
      limit. EKT also solves the problem in which the burst loss of the N
      initial SRTP packets can confuse an SRTP receiver, when the initial RTP
      sequence number is greater than or equal to 2^16 - N.  These features
      can simplify many architectures that implement SRTP.</t>

      <t>EKT provides a way for an SRTP session participant, either a
      sender or a receiver, to securely transport its SRTP master key
      and current SRTP rollover counter to the other participants in
      the session. This data, possibly in conjunction with additional
      data provided by an external signaling protocol, furnishes the
      information needed by the receiver to instantiate an SRTP/SRTCP
      receiver context.</t>

      <t>EKT does not control the manner in which the SSRC and master key are
      generated; it is only concerned with their secure transport. Those
      values may be generated on demand by the SRTP endpoint, or may be
      dictated by an external mechanism such as a signaling agent or a secure
      group controller.</t>

      <t>EKT is not intended to replace external key establishment mechanisms
      such as SDP Security Descriptions <xref target="RFC4568"></xref>,
      DTLS-SRTP <xref target="RFC5764"></xref>, or MIKEY <xref
      target="RFC3830"></xref><xref target="RFC4563"></xref>. Instead, it is
      used in conjunction with those methods, and it relieves them of the
      burden of tightly coordinating every SRTP source among every SRTP
      participant.</t>

      <t>This document is organized as follows. The complete normative
      definition of EKT is contained in <xref
      target="normative"></xref>. It mainly consists of packet processing
      algorithms (<xref target="processing"></xref>) and cryptographic
      definitions (<xref target="cipher"></xref>) . <xref
      target="sdes"></xref>, <xref target="dtls-srtp-kt"></xref>, and
      <xref target="mikey"></xref> define the use of EKT with SDP
      Security Descriptions, DTLS-SRTP, and MIKEY, respectively.
      <xref target="rationale"></xref> provides a design
      rationale. <xref target="interop"></xref> explains how EKT can
      interwork with keying in call signaling.  Security
      Considerations are discussed in <xref target="sec"></xref>, and
      IANA considerations are provided in
      <xref target="iana"></xref>.</t>

      <section title="History">

	<t>
	  RFC Editor Note:  please remove this section prior to publication
	  as an RFC.
	</t>

	<t>
	  This version is substantially revised from earlier versions,
	  in order to make it possible for the EKT data to be removed
	  from a packet without affecting the ability of the receiver
	  to correctly process the data that is present in that
	  packet.  This capability facilitates interoperability
	  between SRTP implementations with different SRTP key
	  management methods.  The changes also greatly simplify the
	  EKT processing rules, and make the EKT data that must be
	  carried in SRTP and/or SRTCP packets somewhat larger.
	</t>
      </section>

      <section title="Conventions Used In This Document">
        <t>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 <xref
        target="RFC2119"></xref>.</t>
      </section>
    </section>

    <section anchor="normative" title="Encrypted Key Transport">
      <t>In EKT, an SRTP master key is encrypted with a key encrypting
      key and the resulting ciphertext is transported in selected
      SRTCP or in selected SRTP packets.  The key encrypting key is
      called an EKT key.  A single such key suffices for a single SRTP
      session, regardless of the number of participants in that
      session.  However, there can be multiple EKT keys used within a
      particular session. </t>

      <t>EKT defines a new method of providing SRTP master keys to an
      endpoint.  In order to convey the ciphertext of the SRTP master
      key, and other additional information, an additional EKT field
      is added to SRTP or SRTCP packets.  When added to SRTCP, the EKT
      field appears at the end of the packet, after the authentication
      tag, if that tag is present, or after the MKI or the SRTCP index
      otherwise.  When added to SRTP, the EKT field appears at the end
      of the packet, after the authentication tag, if that tag is
      present, or after the MKI or the ciphertext of the encrypted
      portion of the packet otherwise.
      </t>

      <section anchor="EKT" title="EKT Field Formats">
        <!--
<t>
We use the following definitions.  An RTP session consists of the RTP
packets sent to a particular destination transport address or set of
such addresses.  An SRTP session is an RTP session protected by SRTP.
A single session participant may have multiple RTP sources.
</t>
-->

        <t>
	  The EKT Field uses one of the two formats defined
	  below. These two formats can always be unambiguously
	  distinguished on receipt by examining the final bit of the
	  EKT Field, which is also the final bit of the SRTP or SRTCP
	  packet.  The first format is the Full EKT Field (or
	  Full_EKT_Field), and the second is the Short EKT Field (or
	  Short_EKT_Field).  The formats are defined as
        <figure anchor="tag_formats"
                title="EKT data formats">
          <artwork align="center"><![CDATA[
     EKT_Plaintext = SRTP_Master_Key || SSRC || ROC || ISN

    EKT_Ciphertext = EKT_Encrypt(EKT_Key, EKT_Plaintext)

      Full_EKT_Field = EKT_Ciphertext || SPI || '1'

     Short_EKT_Field = Reserved || '0'
]]></artwork>
        </figure>
Here || denotes concatenation, and '1' and '0' denote single one and
zero bits, respectively.  These fields and data elements are defined as
follows:
        <list style="hanging">
	  <!--
	  <t hangText="Base Authentication Tag:">This field contains the
            authentication tag computed by the base authentication function.
            The value of this field is used to check the authenticity of the
            packet.</t>
-->

	  <t hangText="EKT_Plaintext:">The data that is input to the EKT 
	  encryption operation.  This data never appears on the wire, and
	  is used only in computations internal to EKT.  
	  </t>

	  <t hangText="EKT_Ciphertext:">The data that is output from
	  the EKT encryption operation, which is performed as as
	  defined in <xref target="cipher"></xref>.  This field is
	  included in SRTP and SRTCP packets when EKT is in use.  The
	  length of this field is variable, and is equal to the
	  ciphertext size N defined in <xref
	  target="cipher"></xref>. Note that the length of the field
	  is inferable from the SPI field, since the particular EKT
	  cipher used by the sender of a packet can be inferred from
	  that field.  </t>

<!-- dwing comment:
Rollover counter (below) is not shown at all in the figure above, whereas
ISN and SPI are shown.  Is that intentional, or an oversight?
-->
            <t hangText="Rollover Counter (ROC):">The length of this field
            is fixed at 32 bits.  It is included in the EKT plaintext,
            but does not appear on the wire.  On the sender side,
            this field is set to the current value of the SRTP
            rollover counter in the SRTP context associated with the
            SSRC in the SRTP or SRTCP packet. 
	    </t>

            <t hangText="Initial Sequence Number (ISN):">The length of
            this field is fixed at 16 bits. It is included the EKT
            plaintext, but does not appear on the wire.  If this
            field is nonzero, then it indicates the RTP sequence
            number of the initial RTP packet that is protected using
            the SRTP master key conveyed (in encrypted form) by the
            EKT Ciphertext field of this packet. If this field is
            zero, it indicates that the initial RTP packet protected
            using the SRTP master key conveyed in this packet
            preceded, or was concurrent with, the last roll-over of
            the RTP sequence number.</t>

            <t hangText="Security Parameter Index (SPI):">The length of this
            field is fixed at 15 bits.  This field is included
	    in SRTP and SRTCP packets when EKT is in use.  It indicates 
            the appropriate
           EKT key and other parameters for the receiver to use
            when processing the packet. It is an "index" into a table of
            possibilities (which are established via signaling or some other
            out-of-band means), much like the IPsec Security Parameter Index
            <xref target="RFC4301"></xref>. The parameters that are identified
            by this field are: <list style="symbols">
                <t>The EKT key used to process the packet.</t>

                <t>The EKT cipher used to process the packet.</t>

                <t>The Secure RTP parameters associated with the SRTP Master
                Key carried by the packet and the SSRC value in the packet.
                Section 8.2. of <xref target="RFC3711"></xref> summarizes the
                parameters defined by that specification.</t>

                <t>The Master Salt associated with the Master Key. (This value
                is part of the parameters mentioned above, but we call it out
                for emphasis.) The Master Salt is communicated separately, via
                signaling, typically along with the EKT key.</t>
              </list>
<!-- dwing comment
Is there an advantage to using two names, "SPI" and "EKT parameter set",
for these same fields?  Seems unnecessarily confusing.  Can we reduce
it to one name?
-->


	      Together, these data elements assoicated with an
	      instance of EKT are called an EKT parameter set.  Within
	      each SRTP session, each distinct EKT parameter set that
	      may be used MUST be associated with a distinct SPI
	      value, to avoid ambiguity. <!-- Each value MUST be
	      associated with a Key Encrypting Key, and MAY be
	      associated with an EKT cipher and an SRTP parameter set.
	      -->
	    </t>
<!--
            <t hangText="Final bit:">This MUST be 1 in the Full EKT
            Tag format, and MUST be 0 in the Short EKT Tag
            format. This flag distinguishes the packet layout between
            these two cases.</t>
-->

<t hangText="Reserved:">MUST be all zeros on transmission, and MUST be
ignored on reception.</t>

          </list>


Examples of the Full_EKT_Field and Short_EKT_Field formats are shown in
(<xref target="tag_format_base"></xref>) and (<xref
target="tag_format_abbreviated"></xref>), respectively.  These figures
show the on-the-wire data.  The Ciphertext field holds encrypted data,
and thus has no apparent inner structure.
        <figure anchor="tag_format_base"
                title="An example of the Full EKT Field format">
          <artwork align="center"><![CDATA[
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
:                                                               :
:                        EKT Ciphertext                         :
:                                                               :
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
|   Security Parameter Index  |1|
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
]]></artwork>
        </figure>

        <figure anchor="tag_format_abbreviated"
                title="An example of the Short EKT Field format">
          <artwork align="center"><![CDATA[
 0                   1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+
|  Reserved   |0|
+-+-+-+-+-+-+-+-+
]]></artwork>
        </figure>
      </t>


<!--
        <t>The Short EKT Field field contains the following sub-fields:
        <list style="hanging">
            <t hangText="Reserved:">7 bits. MUST be 0 on transmission and MUST
            be ignored on reception.</t>

            <t hangText="Final Bit:">This MUST be 0. This flag distinguishes
            the packet layout between<xref target="tag_format_base"></xref> or
            <xref target="tag_format_abbreviated"></xref>.s</t>
          </list></t>
-->
      </section>

      <section anchor="processing" title="Packet Processing and State Machine">
        <t>At any given time, each SRTP/SRTCP source has associated with it a
        single EKT parameter set. This parameter set is used to process all
        outbound packets, and is called the outbound parameter set. There may
        be other EKT parameter sets that are used by other SRTP/SRTCP sources
        in the same session. All of these EKT parameter sets SHOULD be stored
        by all of the participants in an SRTP session, for use in processing
        inbound SRTCP traffic.</t>

        <!--
adds some additional steps to the SRTCP packet processing rules.
-->

<!--
	<t>
	  A receiver processes an EKT_Ciphertext by decrypting it,
	  verifying that the SSRC field in the resulting EKT_plaintext
	  matches the SSRC in the session, and accepting the other
	  fields if the SSRC values match.
	</t>
	<t>
	  The new packet formats are simple: the Full_EKT_Field or the
	  Short_EKT_Field are appended at the tail end of the SRTP packet.  
	</t>

-->

<!--
        <t>We next review SRTP authentication and show how the EKT
        authentication method is built on top of a base authentication method.
        An SRTP or SRTCP authentication method consists of a tag-generation
        function and a verification function. The tag-generation function
        takes as input a secret key, the data to be authenticated, and the
        packet index. It provides an authentication tag as its sole output,
        and is used in the processing of outbound packets. The verification
        function takes as input a secret key, the data to be authenticated,
        the packet index, and the authentication tag. It returns an indication
        of whether or not the data, index, and tag are valid or not. It is
        used in the processing of inbound packets. EKT defines a
        tag-generation function in terms of the base tag-generation function,
        and defines a verification function in terms of the base verification
        function. The tag-generation function is used to process outbound
        packets, and the verification function is used to process inbound
        packets.</t>
-->
        <section anchor="outbound" title="Outbound Processing">
          <t>When an SRTP or SRTCP packet is to be sent, the EKT field
          for that packet is created as follows, or uses an equivalent
          set of steps.  The creation of the EKT field MUST precede
          the normal SRTP or SRTCP packet processing, so that the ROC
          used in EKT is the same as the one used in the SRTP or SRTCP
          processing.
	  </t>


<!-- dwing comment:
I found the paragraph below very difficult to parse.  I made a stab
at what I think was the intended text.

Question:  should we encourage sending the Full EKT twice (or even
three times) to accomodate packet loss?
-->
	  <t>
	  First, the sender decides whether to use the Full or Short
          format.  When sending EKT with SRTP, the Full format SHOULD
          be used on the initial SRTP packet in a session and after
          each rekeying event.  When sending EKT with SRTCP, the Full
          format MUST be used.  Not all SRTP or SRTCP packets need to
          include the EKT key, but it SHOULD be included with some
          regularity, e.g., every second or every ten seconds, though
          it need not be sent on a regular schedule.

	  <!--
  (see <xref target="fig2"/> and
<xref target="fig3"/>).
-->
	  </t>
	  <t>
	    If the Short format is used, an all-zero Reserved octet is appended
	    to the packet.  Otherwise, processing continues as
	    follows.
	  </t>
	  <t>
	    The Rollover Counter field in the packet is set to the
	    current value of the SRTP rollover counter (represented as
	    an unsigned integer in network byte order).</t>

          <t>The Initial Sequence Number field is set to zero, if the
          initial RTP packet protected using the current SRTP master
          key for this source preceded, or was concurrent with, the
          last roll-over of the RTP sequence number. Otherwise, that
          field is set to the value of the RTP sequence number of the
          initial RTP packet that was or will be protected by that
          key. When the SRTP master key corresponding to a source is
          changed, the new key SHOULD be communicated in advance via
          EKT. (Note that the ISN field allows the receiver to know
          when it should start using the new key to process SRTP
          packets.) This enables the rekeying event to be communicated
          before any RTP packets are protected with the new key. The
          rekeying event MUST NOT change the value of ROC (otherwise,
          the current value of the ROC would not be known to late
          joiners of existing sessions).</t>

          <t>The Security Parameter Index field is set to the value of the
          Security Parameter Index that is associated with the outbound
          parameter set.<!-- If only one Key
Encrypting Key is provided for the session, then an all-null value for
the identifier SHOULD be used,
--></t>

	  <t>
	    The EKT_Plaintext field is computed from the SRTP Master Key, SSRC, 
	    ROC, and ISN fields, as shown in <xref target="tag_formats"/>.
	  </t>

          <t>The EKT_Ciphertext field is set to the ciphertext created
          by encrypting the EKT_Plaintext with the EKT cipher, using the KEK
          as the encryption key. The encryption process is detailed in <xref
          target="cipher"></xref>. Implementations MAY cache the value of this
          field to avoid recomputing it for each packet that is sent.<!--
Then the EKT Ciphertext
  field is computed using the EKT cipher's encryption function.  The
  SRTP master key corresponding to the SSRC (and MKI, if present) in
  the packet is the plaintext input, and the Key Encrypting Key is used
  as its key.
   <list style="empty">
     <t> Note: the value of the EKT Ciphertext field is identical
         in successive packets protected by the same KEK and SRTP
         master key.  This value MAY be cached by an SRTP sender to
         minimize computational effort.  (This property follows from
         the fact that the EKT cipher is deterministic, and so
         identical ciphertexts will decrypt to identical plaintexts.
         When the ciphertext form of the master key matches the
         EKT Ciphertext field in a packet, we know that the
         decryption of that field will match the master key.)
     </t>
     </list>
   The computed value of the EKT Ciphertext field is written into
   the packet.
--></t>


<!--
        <section anchor="abbreviated" title="Computing the Short Authentication Field">
<t>
The Short Authentication Tag format consists only of a single byte with
all bits set to zero.  
</t>
</section>

-->

            <!--
<t>
The Initial Sequence Number (ISN) field is set to zero unless the SRTP
master key has been changed, and the EKT fields
are being carried in an SRTCP packet.  In that case, then
that field is set to the SRTP sequence number of the initial packet
that was sent immediately after the current master key was put into
use.
</t>
<t>
The Security Parameter Index field is set to the value of the SPI
that is currently in use.
</t>
<figure anchor="fig2" title="EKT outbound processing; generating the  
EKT Ciphertext">
<artwork>
                             +<<<<< SRTP master key (plaintext)
                             |
                             v
                       ++++++++++++++
                       | Encryption |
                       |  Function  |<, Key Encrypting Key
                       ++++++++++++++
                             |
                             v
                    EKT Ciphertext (ciphertext)
</artwork>
</figure>
<figure anchor="fig3" title="EKT outbound processing; generating the  
Authentication Tag">
<artwork>
            authenticated portion of SRTCP packet
                             |
                             v
                     ++++++++++++++++++       SRTCP
                     |  Base SRTCP    |	 authentication
                     | Tag Generation |<   key
                     ++++++++++++++++++
                             |
                             v
                     Authentication Tag
</artwork>
</figure>
-->
        </section>

        <section anchor="inbound" title="Inbound Processing">
          <t>
	    When an SRTP or SRTCP packet containing and EKT
	    field is received, it is processed as follows, <!-- (see <xref
	    target="fig4"></xref>) --> or uses an equivalent set of steps. 
	    Inbound EKT processing MUST take place prior to 
	    the usual SRTP or SRTCP processing.
	  </t>

          <t><list style="numbers">
	    <t>
	      The final bit is checked to determine which EKT format
	      is in use.  If the packet contains a Short EKT Tag, then
	      the EKT Tag is stripped off of the packet, and then the
	      normal SRTP or SRTCP processing is applied.
	      If the packet contains a Full EKT Tag, then 
	      processing continues as described below.
	    </t>
              <t> 
		The Security Parameter Index (SPI) field is checked to determine
              which EKT parameter set should be used when processing the
              packet. If multiple parameter sets have been defined for the SRTP
              session, then the one that is associated with the value of the SPI
	      field in the packet is used. This parameter set is called the
              matching parameter set below. If there is no matching SPI, then
              the verification function MUST return an indication of
              authentication failure, and the steps described below are not
              performed.<!-- If there is
only a single session key, it SHOULD be used only if the field is set
to the all-null value.
--></t>

	      <t>
		The EKT_Ciphertext is decrypted using the EKT_Key and
		EKT_Cipher in the matching parameter set, as described
		in <xref target="cipher"/>.  If the EKT decryption
		operation returns an authentication failure, then the
<!-- dwing comment:
The text immediately above assumes the "decryption operation" provides
authentication, too.  That's true if running GCM, but not true if
we're running AES with a separate authentication tag.  Can you tweak
the above text to accomodate both styles?  I don't know how to best
make that change.
-->

		packet processing halts with an indication of failure.
		Otherwise, the resulting EKT_Plaintext is parsed as
		described in <xref target="tag_formats"/>, to recover
		the SRTP Master Key, SSRC, ROC, and ISN fields.
	      </t>

	      <t>
		The SSRC field output from the decryption operation is
		compared to the SSRC field from the SRTP header.
		If the values of the two fields do not match, 
		then packet processing halts with an indication
		of failure.   Otherwise, it continues as follows.
	      </t>
	      
	      <t>
		If the ROC from the EKT_Plaintext is less than the ROC
		in the SRTP context, then packet processing halts.
		Otherwise, the ROC in the SRTP context is set to the
		value of the ROC from the EKT_Plaintext, and the SRTP
		Master Key from the EKT_Plaintext is accepted as the
		SRTP master key corresponding to the SRTP source that
		sent the packet. If an MKI is present in the packet,
		then the master key corresponds to the particular SSRC
		and MKI combination. If there is no SRTP crypto
		context corresponding to the SSRC in the packet, then
		a new crypto context is created.  If the crypto
		context is not new, then the rollover counter in the
		context MUST NOT be set to a value lower than its
		current value. (If the replay protection step
		described above is performed, it ensures that this
		requirement is satisfied.)</t>

		<!--

              <t>If there is already an SRTP crypto context associated with
              the SSRC in the packet, and replay protection is in use, then
              the receiver performs the replay check described in Section
              3.3.2 of <xref target="RFC3711"></xref>. If the EKT fields are
              conveyed in an RTCP packet, then the packet index used in that
              check is formed from the Rollover Counter and the Initial
              Sequence Number fields in that packet. If the EKT fields are
              conveyed in an SRTP packet, then the packet index used in that
              check is formed from the EKT Rollover Counter field and the RTP
              Sequence Number in that packet.</t>

	      -->

              <t>If the Initial Sequence Number field is nonzero, then the
              initial sequence number for the SRTP master key is set to the
              packet index created by appending that field to the current
              rollover counter and treating the result as a 48-bit unsigned
              integer. The initial sequence number for the master key is
              equivalent to the "From" value of the <From, To> pair of
              indices (Section 8.1.1 of <xref target="RFC3711"></xref>) that
              can be associated with a master key.</t>

              <t>The newly accepted SRTP master key, the SRTP parameters from
              the matching parameter set, the SSRC from the packet, and the
              MKI from the packet, if one is present, are stored in the crypto
              context associated with the SRTP source. The SRTP Key Derivation
              algorithm is run in order to compute the SRTP encryption and
              authentication keys, and those keys are stored for use in SRTP
              processing of inbound packets. The Key Derivation algorithm
              takes as input the newly accepted SRTP master key, along with
              the Master Salt from the matching parameter set. <list
                  style="empty">
                  <t>Implementation note: the receiver may want to retain old
                  master keys for some brief period of time, so that out of
                  order packets can be processed.</t>
                </list></t>

              <t>At this point, EKT processing has successfully
	      completed, and the normal SRTP or SRTCP processing takes place.<!--
Original text: If the value of the Rollover Counter field (when  
considered as an
unsigned integer in network byte order) is greater than the current
value of the SRTP rollover counter, then the rollover counter is set
to the value of the field.
--> <list style="empty">
                  <t>Implementation note: the value of the EKT
                  Ciphertext field is identical in successive packets
                  protected by the same EKT parameter set and the same
                  SRTP master key and ROC.  This ciphertext value MAY be cached
                  by an SRTP receiver to minimize computational
                  effort by noting when the SRTP
                  master key is unchanged and avoiding repeating
                  Steps 2, 3, 4 5, and 6.</t>
                </list></t>
            </list><!--
If the SSRC field in the packet does correspond to a known SRTP
source, then the packet is processed as follows, or using an
equivalent set of steps.
<list style="numbers">
<t>
The appropriate parameter set is selected using the Security Parameter
Index field, as in Step 1 above, and the EKT Ciphertext
field is decrypted using the Key Encrypting Key as in Step 2 above.
The resulting plaintext is used as a provisional SRTP master key.
</t>
<t>
If the provisional key matches the key associated with the SSRC (and
MKI, if one is present), then the packet is processed as follows.  The
base authentication function is used to check the authenticity of the
packet, as described in Step 3 above.  If the check passes, the packet
is accepted, and then the SRTP rollover counter is set to the value of
the Rollover Counter field.
</t>
<t>
If the provisional key does not match the SRTP master key associated
with the SRTP source (and MKI, if present), then processing proceeds
as follows.  The authenticity of the packet is checked using the
provisional key and base SRTCP authentication function, as in Step 3
above.  If that check passes, then the packet is accepted, and the
context associated with the SRTP source is set as described in Step 5
above.  SRTP master keys can be associated with an initial packet
index (Section 8.1.1. of <xref target="RFC3711"/>).  The initial
packet index associated with the new SRTP master key is formed from
the value of the Initial Sequence Number and Rollover Counter fields
(considered as an unsigned integer in network byte order), by
multiplying the ROC by 2^16 then adding the ISN to the result.
</t>
</list>
--></t>

<!--
          <figure anchor="fig4" title="EKT inbound processing.">
            <artwork align="center"><![CDATA[
             +------- EKT Ciphertext
             |
             v
       +------------+
       | Decryption |
       |  Function  |< ------------------------- Key Encrypting Key
       +------------+
             |                    +----------------+     EKT
    +--------+-- provisional ---->|     SRTCP      |< -  master
    |            master key       | Key Derivation |     salt
    |                             +----------------+
    |                                     |
    |                    provisional SRTCP authentication key
    |                                     |
    |                                     v
    |                             +----------------+
    |   authenticated portion --> |   Base SRTCP   |
    |   authentication tag -----> |  Verification  |
    |                             +----------------+
    |                                     |
    |        +----------------+         +---+
    |        |  return FAIL   |<- FAIL -| ? |
    |        +----------------+         +---+
    |                                     |
    |        +----------------+           |
    +------->| set master key,|<- PASS ---+
             | ROC, and MKI   |
             +----------------+
                     |
                     v
             +----------------+
             |  return PASS   |
             +----------------+
]]></artwork>
          </figure>
-->

        </section>
      </section>

      <section anchor="cipher" title="Ciphers">
        <t>EKT uses an authenticated cipher to encrypt the SRTP master
        keys, ROC, and ISN. We first specify the interface to the
        cipher, in order to abstract the interface away from the
        details of that function. We then define the cipher that is
        used in EKT by default. This cipher MUST be implemented, but
        another cipher that conforms to this interface MAY be used, in
        which case its use MUST be coordinated by external means
        (e.g., call signaling).</t>

        <t>An EKT cipher consists of an encryption function and a decryption
        function. The encryption function E(K, P) takes the following inputs:
        <list style="symbols">
            <t>a secret key K with a length of L bytes, and</t>

            <t>a plaintext value P with a length of M bytes.</t>
          </list> The encryption function returns a ciphertext value C whose
        length is N bytes, where N is at least M.  The decryption function D(K,
        C) takes the following inputs: <list style="symbols">
            <t>a secret key K with a length of L bytes, and</t>

            <t>a ciphertext value C with a length of N bytes.</t>
            </list> The decryption function returns a plaintext value
            P that is M bytes long, or returns an indication that the
            decryption operation failed because the ciphertext was
            invalid (i.e. it was not generated by the encryption of
            plaintext with the key K). 
	    </t>
	    <t>
	      These functions have the
            property that D(K, E(K, P)) = P for all values of K and P.
            Each cipher also has a limit T on the number of times that
            it can be used with any fixed key value.  For each key,
            the encryption function MUST NOT be invoked on more than T
            distinct values of P, and the decryption function MUST NOT
            be invoked on more than T distinct values of C.</t>

	<t>The length of the EKT Plaintext is ten bytes, plus the length 
	of the SRTP Master Key.</t>

	<t>Security requirements for EKT ciphers are discussed in <xref target="sec"/>. </t>

        <section anchor="DefaultCipher" title="The Default Cipher">
          <t>
	    The default EKT Cipher is the Advanced Encryption Standard
	    (AES) <xref target="FIPS197"/> Key Wrap with Padding <xref
	    target="RFC5649"></xref> algorithm, which can be used with
	    plaintexts larger than 16 bytes in length, and is thus
	    suitable for keys of any size.  It requires a plaintext
	    length M that is at least eight bytes, and it returns
	    a ciphertext with a length of N = M + 8 bytes.  It can be
	    used with key sizes of L = 16, 24, and 32, and its use
	    with those key sizes is indicated as AESKW_128, AESKW_192,
	    and AESKW_256, respectively.  The key size determines the
	    length of the AES key used by the Key Wrap algorithm.
	    With this cipher, T=2^48.
	  </t>
	  <t>
	    When AES-128 is used in SRTP and/or SRTCP, AESKW_128
	    SHOULD be used in EKT.  In this case, the EKT Plaintext is
	    26 bytes long, the EKT Ciphertext is 40 bytes long, and
	    the Full EKT field is 42 bytes long.
	  </t>
	  <t>
	    When AES-192 is used in SRTP and/or SRTCP, AESKW_192
	    SHOULD be used in EKT.  In this case, the EKT Plaintext is
	    34 bytes long, the EKT Ciphertext is 48 bytes long, and
	    the Full EKT field is 50 bytes long.
	  </t>
<!-- dwing comment:
the lengths above appear to be different from the lengths at
dwing-comment-length (search for string dwing-comment-length)
-->

	  <t>
	    When AES-256 is used in SRTP and/or SRTCP, AESKW_256
	    SHOULD be used in EKT.  In this case, the EKT Plaintext is
	    42 bytes long, the EKT Ciphertext is 56 bytes long, and
	    the Full EKT field is 58 bytes long.
	  </t>
<!--
The default cipher is the Advanced Encryption Standard (AES)
          <xref target="FIPS197"></xref> with 128-bit keys, in Electronic
          Codebook (ECB) Mode. Its parameters are fixed at L=16, M=16, and
          T=2^48. Note that M matches the size of the SRTP master keys used by
          the default SRTP key derivation function; if an SRTP cipher with a
          different SRTP master key length is to be used with EKT, then
          another EKT cipher must be used. ECB is the simplest mode of
          operation of a block cipher, in which the block cipher is used in
          its raw form.</t>
-->
        </section>

<!--
	<section title="AES ECB">
	  <t>
	    The simplest EKT cipher is the Advanced Encryption
	    Standard (AES)
          <xref target="FIPS197"></xref> with 128-bit keys, in
          Electronic Codebook (ECB) Mode. Its use is indicated as
          AES_ECB, and its parameters are fixed at L=16, M=16, and
          T=2^48. Note that M matches the size of the SRTP master keys
          used by the default SRTP key derivation function; if an SRTP
          cipher with a different SRTP master key length is to be used
          with EKT, then another EKT cipher must be used. ECB is the
          simplest mode of operation of a block cipher, in which the
          block cipher is used in its raw form.
	    </t>
	  </section>
   

-->
<!--
     <section anchor="AESKeyWrap" title="The AES Key Wrap Cipher">
          <t>The AES Key Wrap  defines a cipher
          that can be used with plaintexts larger than 16 bytes in length. It
          requires a plaintext length M that is a multiple of eight bytes, and
          it returns a ciphertext with a length of N = M + 8 bytes. It can be
          used with key sizes of L = 16, 24, and 32. The key size determines
          the length of the AES key used by the Key Wrap algorithm. With this
          cipher, T=2^48.</t>
        </section>
-->
        <section title="Other EKT Ciphers">
          <t>Other specifications may extend this one by defining
          other EKT ciphers per <xref target="iana"></xref>. This
          section defines how those ciphers interact with this
          specification.</t>

          <t>An EKT cipher determines how the EKT Ciphertext field is
          written, and how it is processed when it is read. This field is
          opaque to the other aspects of EKT processing. EKT ciphers are free
          to use this field in any way, but they SHOULD NOT use other EKT or
          SRTP fields as an input.  The values of the parameters L, M, N, and T
          MUST be defined by each EKT cipher, and those values MUST be
          inferable from the EKT parameter set.</t>
        </section>
      </section>

      <section anchor="SynchronizingOperation" title="Synchronizing Operation">
        <t>A participant in a session MAY opt to use a particular EKT key to
        protect outbound packets after it accepts that EKT key for protecting
        inbound traffic. In this case, the fact that one participant has
        changed to using a new EKT key for outbound traffic can trigger other
        participants to switch to using the same key.</t>

        <t>An SRTP/SRTCP source SHOULD change its SRTP master key after its
        EKT key has been changed. This will ensure that the set of
        participants able to decrypt the traffic will be limited to those who
        know the current EKT key.</t>

        <t>EKT can be transported over SRTCP, but some of the
        information that it conveys is used for SRTP processing; some
        elements of the EKT parameter set apply to both SRTP and
        SRTCP. Furthermore, SRTCP packets can be lost and both SRTP
        and SRTCP packets may be delivered out of order. This can lead
        to various race conditions if EKT is transported over SRTCP
        but not SRTP, which we review below.</t>

        <t>When joining an SRTP session, SRTP packets may be received before
        any EKT over SRTCP packets, which implies the crypto context has not
        been established, unless other external signaling mechanism has done
        so. Rather than automatically discarding such SRTP packets, the
        receiver MAY want to provisionally place them in a jitter buffer and
        delay discarding them until playout time.</t>

        <t>When an SRTP source using EKT over SRTCP performs a rekeying
        operation, there is a race between the actual rekeying signaled via
        SRTCP and the SRTP packets secured by the new keying material. If the
        SRTP packets are received first, they will fail authentication;
        alternatively, if authentication is not being used, they will decrypt
        to unintelligible random-looking plaintext. (Note, however, that <xref
        target="RFC3711"></xref> says that SRTP "SHOULD NOT be used without
        message authentication".) In order to address this problem, the
        rekeying event can be sent before packets using the new SRTP master
        key are sent (by use of the ISN field). Another solution involves
        using an MKI at the expense of added overhead in each SRTP packet.
        Alternatively, receivers MAY want to delay discarding packets from
        known SSRCs that fail authentication in anticipation of receiving a
        rekeying event via EKT (SRTCP) shortly.</t>

        <t><!--
<list style="empty">
<t>
Note: It was considered to include <From, To> functionality in EKT to
indicate when a particular set of parameters would apply. However, if
this is to be done without increasing the SRTP packet size, then the
 From and To values can only include the 16-bit SRTP Sequence Number,
not the full (32-bit) SRTP Packet Index (which includes the
ROC). Thus, once the sequence number wraps, such an indication will no
longer be correct.
</t>
</list>
--></t>

        <t>The ROC signaled via EKT over SRTCP may be off by one when it is
        received by the other party(ies) in the session. In order to deal with
        this, receivers should simply follow the SRTP packet index estimation
        procedures defined in <xref target="RFC3711">Section 3.3.1</xref>.</t>
      </section>

      <section anchor="srtp" title="Transport">
        <t>EKT MUST be used over SRTCP, whenever RTCP is in use. EKT MAY be
        used over SRTP. When EKT over SRTP is used in an SRTP session in which
        SRTCP is available, then EKT MUST be used for both SRTP and SRTCP.</t>

        <t>The packet processing, state machine, and Authentication Tag format
        for EKT over SRTP are nearly identical to that for EKT over SRTCP.
        Differences are highlighted in <xref target="outbound"></xref> and
        <xref target="inbound"></xref>.</t>
      </section>

      <section title="Timing and Reliability Consideration">
        <t>SRTCP communicates the master key and ROC for the SRTP
        session.  Thus, as explained above, if SRTP packets are
        received prior to the corresponding SRTCP (EKT) packet, a race
        condition occurs. From an EKT point of view, it is therefore
        desirable for an SRTP sender to send an EKT packet containing
        the Base Authentication Tag as soon as possible, and in no
        case any later than when the initial SRTP packet is sent. It
        is RECOMMENDED that the Base Authentication Tag be transmitted
        3 times (to accomodate packet loss) and to provide a reliable
        indication to the receiver that the sender is now using the
        EKT key.  If the Base Authentication Tag sent in SRTCP, the
        SRTCP timing rules associated with the profile under which it
        runs (e.g., RTP/SAVP or RTP/SAVPF) MUST be obeyed.  Subject to
        that constraint, SRTP senders using EKT over SRTCP SHOULD send
        an SRTCP packet as soon as possible after joining a
        session. Note that there is no need for SRTP receivers to do
        so. Also note, that per RFC 3550, Section 6.2, it is
        permissible to send a compound RTCP packet immediately after
        joining a unicast session (but not a multicast session).</t>

        <t>SRTCP is not reliable and hence SRTCP packets may be lost. This is
        obviously a problem for endpoints joining an SRTP session and
        receiving SRTP traffic (as opposed to SRTCP), or for endpoints
        receiving SRTP traffic following a rekeying event. To reduce the
        impact of lost packets, SRTP senders using EKT over SRTCP SHOULD send
        SRTCP packets as often as allowed by the profile under which they
        operate.<!-- Editor's
Note: For unicast sessions, use of a reliable protocol to transport
EKT packets should be considered. STUN (as used by ICE) is one option,
however we would need to include SSRC identification as well as
procedures for SSRC collission detection and resolution then.
--></t>
      </section>
    </section>

    <section anchor="sdes" title="Use of EKT with SDP Security Descriptions">
      <t>The SDP Security Descriptions (SDESC) <xref target="RFC4568"></xref>
      specification defines a generic framework for negotiating security
      parameters for media streams negotiated via the Session Description
      Protocol by use of a new SDP "crypto" attribute and the Offer/Answer
      procedures defined in <xref target="RFC3264"></xref>. In addition to the
      general framework, SDES also defines how to use that framework
      specifically to negotiate security parameters for Secure RTP. Below, we
      first provide a brief recap of the crypto attribute when used for SRTP
      and we then explain how it is complementary to EKT. In the rest of this
      Section, we provide extensions to the crypto attribute and associated
      offer/answer procedures to define its use with EKT.</t>

      <section title="SDP Security Descriptions Recap">
        <t>The SRTP crypto attribute defined for SDESC contains a tag followed
        by three types of parameters (refer to <xref target="RFC4568"></xref>
        for details): <list style="symbols">
            <t>Crypto-suite. Identifies the encryption and authentication
            transform</t>

            <t>Key parameters. SRTP keying material and parameters.</t>

            <t>Session parameters. Additional (optional) SRTP parameters such
            as Key Derivation Rate, Forward Error Correction Order, use of
            unencrypted SRTP, and other parameters defined by SDESC.</t>
          </list> The crypto attributes in the example SDP in <xref
        target="figSDP"></xref> illustrate these parameters.</t>

        <figure anchor="figSDP" title="SDP Security Descriptions example">
          <artwork align="center"><![CDATA[
v=0
o=sam 2890844526 2890842807 IN IP4 192.0.2.5
s=SRTP Discussion
i=A discussion of Secure RTP
u=http://www.example.com/seminars/srtp.pdf
e=marge@example.com (Marge Simpson)
c=IN IP4 192.0.2.12
t=2873397496 2873404696
m=audio 49170 RTP/SAVP 0
a=crypto:1 AES_CM_128_HMAC_SHA1_80
   inline:WVNfX19zZW1jdGwgKCkgewkyMjA7fQp9CnVubGVz|2^20|1:4
   FEC_ORDER=FEC_SRTP
a=crypto:2 F8_128_HMAC_SHA1_80
   inline:MTIzNDU2Nzg5QUJDREUwMTIzNDU2Nzg5QUJjZGVm|2^20|1:4;
   inline:QUJjZGVmMTIzNDU2Nzg5QUJDREUwMTIzNDU2Nzg5|2^20|2:4
   FEC_ORDER=FEC_SRTP
]]></artwork>

          <postamble></postamble>
        </figure>

        <t>For legibility the SDP shows line breaks that are not present on
        the wire.</t>

        <t>The first crypto attribute has the tag "1" and uses the
        crypto-suite AES_CM_128_HMAC_SHA1_80. The "inline" parameter provides
        the SRTP master key and salt, the master key lifetime (number of
        packets), and the (optional) Master Key Identifier (MKI) whose value
        is "1" and has a byte length of "4" in the SRTP packets. Finally, the
        FEC_ORDER session parameter indicates the order of Forward Error
        Correction used (FEC is applied before SRTP processing by the sender
        of the SRTP media).</t>

        <!-- EDITOR's NOTE: EKT only includes the master key. What
	about the salt ? -->

        <t>The second crypto attribute has the tag "2" and uses the
        crypto-suite F8_128_HMAC_SHA1_80. It includes two SRTP master keys and
        associated salts. The first one is used with the MKI value 1, whereas
        the second one is used with the MKI value 2. Finally, the FEC_ORDER
        session parameter indicates the order of Forward Error Correction
        used.</t>
      </section>

      <section title="Relationship between EKT and SDP Security Descriptions">
        <t>SDP Security Descriptions <xref target="RFC4568"></xref> define a
        generic framework for negotiating security parameters for media
        streams negotiated via the Session Description Protocol by use of the
        Offer/Answer procedures defined in <xref target="RFC3264"></xref>. In
        addition to the general framework, SDESC also defines how to use it
        specifically to negotiate security parameters for Secure RTP.</t>

        <t>EKT and SDESC are complementary. SDESC can negotiate several of the
        SRTP security parameters (e.g., cipher and use of Master Key
        Identifier/MKI) as well as SRTP master keys. SDESC, however, does not
        negotiate SSRCs and their associated Rollover Counter (ROC). Instead,
        SDESC relies on a so-called "late binding", where a newly observed
        SSRC will have its crypto context initialized to a ROC value of zero.
        Clearly, this does not work for participants joining an SRTP session
        that has been established for a while and hence has a non-zero ROC. It
        is impossible to use SDESC to join an SRTP session that is already in
        progress. In this case, EKT on the endpoint running SDP Security can
        provide the additional signaling necessary to communicate the ROC
        (Section 6.4.1 of <xref target="RFC4568"></xref>). The use of EKT
        solves this problem by communicating the ROC associated with the SSRC
        in the media plane.</t>

        <t>SDP Security Descriptions negotiates different SRTP master keys in
        the send and receive direction. The offer contains the master key used
        by the offerer to send media, and the answer contains the master key
        used by the answerer to send media. Consequently, if media is received
        by the offerer prior to the answer being received, the offerer does
        not know the master key being used. Use of SDP security preconditions
        can solve this problem, however it requires an additional round-trip
        as well as a more complicated state machine. EKT solves this problem
        by simply sending the master key used in the media plane thereby
        avoiding the need for security preconditions.</t>

        <t>If multiple crypto-suites were offered, the offerer also will not
        know which of the crypto-suites offered was selected until the answer
        is received. EKT solves this problem by using a correlator, the
        Security Parameter Index (SPI), which uniquely identifies each crypto
        attribute in the offer.</t>

        <!--
EDITOR'S NOTE: How does EKT help solve that ? One solution would be to
include the "tag" from the selected crypto attribute, however this is
sdes specific (may need something else for MIKEY). It also does not
solve the issue of declarative parameters included in the answer,
which the offerer needs to know about to correctly process incoming
media. Propose to simply note that as a limitation of the solution (if
you want the whole nine yards, you need security preconditions).
  -->

        <t>One of the primary call signaling protocols using offer/answer is
        the Session Initiation Protocol (SIP) <xref target="RFC3261"></xref>.
        SIP uses the INVITE message to initiate a media session and typically
        includes an offer SDP in the INVITE. An INVITE may be "forked" to
        multiple recipients which potentially can lead to multiple answers
        being received. SDESC, however, does not properly support this
        scenario, mainly because SDP and RTP/RTCP does not contain sufficient
        information to allow for correlation of an incoming RTP/RTCP packet
        with a particular answer SDP. Note that extensions providing this
        correlation do exist (e.g., Interactive Connectivity Establishment
        (ICE)). SDESC addresses this point-to-multipoint problem by moving
        each answer to a separate RTP transport address thereby turning a
        point-to-multipoint scenario into multiple point-to-point scenarios.
        There are however significant disadvantages to doing so. As long as
        the crypto attribute in the answer does not contain any declarative
        parameters that differ from those in the offer, EKT solves this
        problem by use of the SPI correlator and communication of the
        answerer's SRTP master key in EKT.</t>

        <!--
EDITOR's NOTE: How does EKT help solve that ? We would need to include
a correlator that is found in both the SDP and the RTCP packet. It
should be noted that ICE already provides this.
-->

        <t>As can be seen from the above, the combination of EKT and SDESC
        provides a better solution to SRTP negotiation for offer/answer than
        either of them alone. SDESC negotiates the various SRTP crypto
        parameters (which EKT does not), whereas EKT addresses the
        shortcomings of SDESC.</t>
      </section>

      <section title="Overview of Combined EKT and SDP Security Description Operation">
        <t>We define three session extension parameters to SDESC to
        communicate the EKT cipher, EKT key, and Security Parameter Index to
        the peer. The original SDESC parameters are used as defined in <xref
        target="RFC4568"></xref>, however the procedures associated with the
        SRTP master key differ slightly, since both SDESC and EKT communicate
        an SRTP master key. In particular, the SRTP master key communicated
        via SDESC is used only if there is currently no crypto context
        established for the SSRC in question. This will be the case when an
        entity has received only the offer or answer, but has yet to receive a
        valid EKT message from the peer. Once a valid EKT message is received
        for the SSRC, the crypto context is initialized accordingly, and the
        SRTP master key will then be derived from the EKT message. Subsequent
        offer/answer exchanges do not change this: The most recent SRTP master
        key negotiated via EKT will be used, or, if none is available for the
        SSRC in question, the most recent SRTP master key negotiated via
        offer/answer will be used. Note that with these rules, once a valid
        EKT message has been received for a given SSRC, rekeying for that SSRC
        can only be done via EKT. The associated SRTP crypto parameters
        however can be changed via SDESC.<!-- DAM - have not edited this section, shouldn't use default key --></t>
      </section>

      <section anchor="ext"
               title="EKT Extensions to SDP Security Descriptions">
        <t>In order to use EKT and SDESC in conjunction with each other, the
        following new SDES session parameters are defined. These MUST NOT
        appear more than once in a given crypto attribute: <list
            style="hanging">
            <t hangText="EKT_Cipher:">The EKT cipher used to encrypt the SRTP
            Master Key</t>

            <t hangText="EKT_Key:">The EKT key used to encrypt the SRTP Master
            Key</t>

            <t hangText="EKT_SPI:">The EKT Security Parameter Index</t>
          </list></t>

        <t>Below are details on each of these attributes.</t>

        <section title="EKT_Cipher">
          <t>The (optional) EKT_Cipher parameter defines the EKT
          cipher used to encrypt the EKT key with in SRTCP
          packets. The default value is "AESKW_128" in accordance
          with <xref target="DefaultCipher"></xref>.  For the AES Key
          Wrap cipher, the values "AESKW_128", "AESKW_192", and
          "AESKW_256" are defined for values of L=16, 24, and 32
          respectively.  <!-- For the AES ECB cipher, "AES_ECB" is defined.-->
          In the Offer/Answer model, the EKT_Cipher parameter is a
          negotiated parameter.</t>
        </section>

        <section title="EKT_Key">
          <t>The (mandatory) EKT_Key parameter is the key K used to encrypt
          the SRTP Master Key in SRTCP packets. The value is base64 encoded as
          described in Section 4 <xref target="RFC4648"></xref>. When base64
          decoding the key, padding characters (i.e., one or two "=" at the
          end of the base64 encoded data) are discarded (see <xref
          target="RFC4648"></xref> for details). Base64 encoding assumes that
          the base64 encoding input is an integral number of octets. If a
          given EKT cipher requires the use of a key with a length that is not
          an integral number of octets, said cipher MUST define a padding
          scheme that results in the base64 input being an integral number of
          octets. For example, if the length defined was 250 bits, then 6
          padding bits would be needed, which could be defined to be the last
          6 bits in a 256 bit input. In the Offer/Answer model, the EKT_Key
          parameter is a negotiated parameter.</t>
        </section>

        <section title="EKT_SPI">
          <t>The (mandatory) EKT_SPI parameter is the Security Parameter
          Index. It is encoded as an ASCII string representing the hexadecimal
          value of the Security Parameter Index. The SPI identifies the
          *offer* crypto attribute (including the EKT Key and Cipher) being
          used for the associated SRTP session. A crypto attribute corresponds
          to an EKT Parameter Set and hence the SPI effectively identifies a
          particular EKT parameter set. Note that the scope of the SPI is the
          SRTP session, which may or may not be limited to the scope of the
          associated SIP dialog. In particular, if one of the participants in
          an SRTP session is an SRTP translator, the scope of the SRTP session
          is not limited to the scope of a single SIP dialog. However, if all
          of the participants in the session are endpoints or mixers, the
          scope of the SRTP session will correspond to a single SIP dialog. In
          the Offer/Answer model, the EKT_SPI parameter is a negotiated
          parameter.</t>
        </section>
      </section>

      <section title="Offer/Answer Procedures">
        <t>In this section, we provide the offer/answer procedures associated
        with use of the three new SDESC parameters defined in Section <xref
        target="ext"></xref>. Since SDESC is defined only for unicast streams,
        we provide only offer/answer procedures for unicast streams here as
        well.</t>

        <section title="Generating the Initial Offer - Unicast Streams">
          <t>When the initial offer is generated, the offerer MUST follow the
          steps defined in <xref target="RFC4568"></xref> Section 7.1.1 as
          well as the following steps.</t>

          <t>For each unicast media line using SDESC and where use of EKT is
          desired, the offerer MUST include one EKT_Key parameter and one
          EKT_SPI parameter in at least one "crypto" attribute (see <xref
          target="RFC4568"></xref>). The EKT_SPI parameter serves to identify
          the EKT parameter set used for a particular SRTCP packet.
          Consequently, within a single media line, a given EKT_SPI value MUST
          NOT be used with multiple crypto attributes. Note that the EKT
          parameter set to use for the session is not yet established at this
          point; each offered crypto attribute contains a candidate EKT
          parameter set. <!--
Can we be more precise in describing a "crypto attribute" -it's an SDES
thing, I assume.
FSA: Right.
--> Furthermore, if the media line refers to an existing SRTP session, then
          any SPI values used for EKT parameter sets in that session MUST NOT
          be remapped to any different EKT parameter sets. When an offer
          describes an SRTP session that is already in progress, the offer
          SHOULD use an EKT parameter set (incl. EKT_SPI and EKT_KEY) that is
          already in use.</t>

          <t>If an EKT_Cipher other than the default cipher is to be used,
          then the EKT_Cipher parameter MUST be included as well.</t>

          <t>If a given crypto attribute includes more than one set of SRTP
          key parameters (SRTP master key, salt, lifetime, MKI), they MUST all
          use the same salt. (EKT requires a single shared salt between all
          the participants in the direct SRTP session).</t>

          <!--
	EDITOR's NOTE: Do we ever see a need to allow for a remapping
	? We really just need to make sure there is no ambiguity, and
	simply cycling the value once we hit 2^32 would be more than
	enough here.
-->

          <!--
	EDITOR'S NOTE: Do we need the ability to negotiate different
	EKT Ciphers in a single O/A exchange ? If so, we will need a
	way for EKT to tell us which one was actually chosen. For now,
	no such capability has been included.
-->

          <t><list style="hanging">
              <t hangText="Important Note:">The scope of the offer/answer
              exchange is the SIP dialog(s) established as a result of the
              INVITE, however the scope of EKT is the direct SRTP session,
              i.e., all the participants that are able to receive SRTP and
              SRTCP packets directly. If an SRTP session spans multiple SIP
              dialogs, the EKT parameter sets MUST be synchronized between all
              the SIP dialogs where SRTP and SRTCP packets can be exchanged.
              In the case where the SIP entity operates as an RTP mixer (and
              hence re-originates SRTP and SRTCP packets with its own SSRC),
              this is not an issue, unless the mixer receives traffic from the
              various participants on the same destination IP address and
              port, in which case further coordination of SPI values and
              crypto parameters may be needed between the SIP dialogs (note
              that SIP forking with multiple early media senders is an example
              of this). However if it operates as an RTP translator,
              synchronized negotiation of the EKT parameter sets on *all* the
              involved SIP dialogs will be needed. This is non-trivial in a
              variety of use cases, and hence use of the combined SDES/EKT
              mechanism with RTP translators should be considered very
              carefully. It should be noted, that use of SRTP with RTP
              translators in general should be considered very carefully as
              well.</t>
            </list></t>

          <!-- Editor's Note: Need to dedicate a section to more fully explain
(and exemplify) the issues here. Need to clarify RTP mixer and RTP
translator operation. Also need to explain SIP forking with early
media from multiple entities.
-->

          <!--
	EDITOR's NOTE: Can we have SRTP session where some
	participants support EKT, and some do not ? For now, I assume
	the answer is no.
-->

          <t>The EKT session parameters can either be included as optional or
          mandatory parameters, however within a given crypto attribute, they
          MUST all be either optional or mandatory.</t>
        </section>

        <section title="Generating the Initial Answer - Unicast Streams">
          <t>When the initial answer is generated, the answerer MUST follow
          the steps defined in <xref target="RFC4568"></xref> Section 7.1.2 as
          well as the following steps.</t>

          <t>For each unicast media line using SDESC, the answerer examines
          the associated crypto attribute(s) for the presence of EKT
          parameters. If mandatory EKT parameters are included with a "crypto"
          attribute, the answerer MUST support those parameters in order to
          accept that offered crypto attribute. If optional EKT parameters are
          included instead, the answerer MAY accept the offered crypto
          attribute without using EKT. However, doing so will prevent the
          offerer from processing any packets received before the answer. If
          neither optional nor mandatory EKT parameters are included with a
          crypto attribute, and that crypto attribute is accepted in the
          answer, EKT MUST NOT be used. If a given a crypto attribute includes
          a mixture of optional and mandatory EKT parameters, or an incomplete
          set of mandatory EKT parameters, that crypto attribute MUST be
          considered invalid.</t>

          <t>When EKT is used with SDESC, the offerer and answerer MUST use
          the same SRTP master salt. Thus, the SRTP key parameter(s) in the
          answer crypto attribute MUST use the same master salt as the one
          accepted from the offer.<!-- DAM - should we say "master salt" instead of just "salt"? --></t>

          <t>When the answerer accepts the offered media line and EKT is being
          used, the crypto attribute included in the answer MUST include the
          same EKT parameter values as found in the accepted crypto attribute
          from the offer (however, if the default EKT cipher is being used, it
          may be omitted). Furthermore, the EKT parameters included MUST be
          mandatory (i.e., no "-" prefix).</t>

          <t>Acceptance of a crypto attribute with EKT parameters leads to
          establishment of the EKT parameter set for the corresponding SRTP
          session. Consequently, the answerer MUST send packets in accordance
          with that particular EKT parameter set only. If the answerer wants
          to enable the offerer to process SRTP packets received by the
          offerer before it receives the answer, the answerer MUST NOT include
          any declarative session parameters that either were not present in
          the offered crypto attribute, or were present but with a different
          value. Otherwise, the offerer's view of the EKT parameter set would
          differ from the answerer's until the answer is received. Similarly,
          unless the offerer and answerer has other means for correlating an
          answer with a particular SRTP session, the answer SHOULD NOT include
          any declarative session parameters that either were not present in
          the offered crypto attribute, or were present but with a different
          value. If this recommendation is not followed and the offerer
          receives multiple answers (e.g., due to SIP forking), the offerer
          may not be able to process incoming media stream packets
          correctly.</t>

          <!--
	EDITOR'S NOTE: It would be sufficient to just include the
	EKT_SPI as there is enough context to determine what was
	accepted otherwise. However, sdes has a rule saying negotiated
	session parameters need to be included in the answer.
-->
        </section>

        <section title="Processing of the Initial Answer - Unicast Streams">
          <t>When the offerer receives the answer, it MUST perform the steps
          in <xref target="RFC4568"></xref> Section 7.1.3 as well as the
          following steps for each SRTP media stream it offered with one or
          more crypto lines containing EKT parameters in it.</t>

          <t>If the answer crypto line contains EKT parameters, and the
          corresponding crypto line from the offer contained the same EKT
          values, use of EKT has been negotiated successfully and MUST be used
          for the media stream. When determining whether the values match,
          optional and mandatory parameters MUST be considered equal.
          Furthermore, if the default EKT cipher is being used, it MAY be
          either present or absent in the offer and/or answer.</t>

          <t>If the answer crypto line does not contain EKT parameters, then
          EKT MUST NOT be used for the corresponding SRTP session. Note that
          if the accepted crypto attribute contained mandatory EKT parameters
          in the offer, and the crypto attribute in the answer does not
          contain EKT parameters, then negotiation has failed (<xref
          target="RFC4568">Section 5.1.3 of</xref>).</t>

          <t>If the answer crypto line contains EKT parameters but the
          corresponding offered crypto line did not, or if the parameters
          don't match or are invalid, then the offerer MUST consider the
          crypto line invalid (see <xref target="RFC4568">Section 7.1.3
          of</xref> for further operation).</t>

          <t>The EKT parameter set is established when the answer is received,
          however there are a couple of special cases to consider here. First
          of all, if an SRTCP packet is received prior to the answer, then the
          EKT parameter set is established provisionally based on the SPI
          included. Once the answer (which may include declarative session
          parameters) is received, the EKT parameter set is fully established.
          The second case involves receipt of multiple answers due to SIP
          forking. In this case, there will be multiple EKT parameter sets;
          one for each SRTP session. As mentioned earlier, reliable
          correlation of SIP dialogs to SRTP sessions requires extensions, and
          hence if one or more of the answers include declarative session
          parameters, it may be difficult to fully establish the EKT parameter
          set for each SRTP session. In the absence of a specific correlation
          mechanism, it is RECOMMENDED, that such correlation be done based on
          the signaled receive IP-address in the SDP and the observed source
          IP-address in incoming SRTP/SRTCP packets, and, if necessary, the
          signaled receive UDP port and the observed source UDP port.</t>
        </section>
      </section>

      <section title="SRTP-Specific Use Outside Offer/Answer">
        <t>Security Descriptions use for SRTP is not defined outside
        offer/answer and hence neither does Security Descriptions with
        EKT.</t>
      </section>

      <section title="Modifying the Session">
        <t>When a media stream using the SRTP security descriptions has been
        established, and a new offer/answer exchange is performed, the offerer
        and answerer MUST follow the steps in <xref target="RFC4568">Section
        7.1.4 of</xref> as well as the following steps. SDESC allows for all
        parameters of the session to be modified, and the EKT session
        parameters are no exception to that, however, there are a few
        additional rules to be adhered to when using EKT.</t>

        <t>It is permissible to start a session without the use of EKT, and
        then subsequently start using EKT, however the converse is not. Thus,
        once use of EKT has been negotiated on a particular media stream, EKT
        MUST continue to be used on that media stream in all subsequent
        offer/answer exchanges.</t>

        <t>The reason for this is that both SDESC and EKT communicate the SRTP
        Master Key with EKT Master Keys taking precedence. Reverting back to
        an SDESC-controlled master key in a synchronized manner is
        difficult.</t>

        <t>Once EKT is being used, the salt for the direct SRTP session MUST
        NOT be changed. Thus, a new offer/answer which does not create a new
        SRTP session (e.g., because it reuses the same IP address and port)
        MUST use the same salt for all crypto attributes as is currently used
        for the direct SRTP session.</t>

        <t>Finally, subsequent offer/answer exchanges MUST NOT remap a given
        SPI value to a different EKT parameter set until 2^32 other mappings
        have been used within the SRTP session. In practice, this requirements
        is most easily met by using a monotonically increasing SPI value
        (modulo 2^32 and starting with zero) per direct SRTP session. Note
        that a direct SRTP session may span multiple SIP dialogs, and in such
        cases coordination of SPI values across those SIP dialogs will be
        required. In the simple point-to-point unicast case without
        translators, the requirement simply applies within each media line in
        the SDP. In the point-to-multipoint case, the requirement applies
        across all the associated SIP dialogs.</t>

        <!--
	EDITOR'S NOTE: I'm not entirely happy and comfortable with
	this restriction. Should probably take another look and see if
	we could accommodate this.
-->
      </section>

      <!--
<section title="Offer/Answer Example">
<t>
To be provided.
</t>
</section>
-->

      <section title="Backwards Compatibility Considerations">
        <t>Backwards compatibility can be achieved in a couple of ways. First
        of all, SDESC allows for session parameters to be prefixed with "-" to
        indicate that they are optional. If the answerer does not support the
        EKT session parameters, such optional parameters will simply be
        ignored. When the answer is received, absence of the parameters will
        indicate that EKT is not being used. Receipt of SRTCP packets prior to
        receipt of such an answer will obviously be problematic (as is
        normally the case for SDESC without EKT).</t>

        <t>Alternatively, SDESC allows for multiple crypto lines to be
        included for a particular media stream. Thus, two crypto lines that
        differ in their use of EKT parameters (presence in one, absence in the
        other) can be used as a way to negotiate use of EKT. When the answer
        is received, the accepted crypto attribute will indicate whether EKT
        is being used or not.</t>
      </section>

      <section title="Grammar">
        <t>The <xref target="RFC5234">ABNF</xref> syntax for the one
        new SDP Security Descriptions session parameter, EKT,
        comprising three parts is shown
        in <xref target="figABNF"></xref>.</t>

        <figure anchor="figABNF" title="ABNF for the EKT session parameters">
          <artwork align="center"><![CDATA[
ekt     = "EKT=" cipher "|" key "|" spi
cipher  = cipher-extension / "AES_128" / "AESKW_128" /
         "AESKW_192" / "AESKW_256"
cipher-extension = 1*(ALPHA / DIGIT / "_")
key     = 1*(base64)    ; See Section 4 of [RFC4648]
base64  = ALPHA / DIGIT / "+" / "/" / "="
spi     = 4HEXDIG   ; See [RFC5234]
]]></artwork>
        </figure>

<!-- dwing comment:
To offer EKT with both AES256 and (let's pretend) AES2048, would we do
this with separate a=crypto lines, or with multiple EKT= on the same
a=crypto line?  Should we explain this?
-->

        <t>Using the example from <xref target="figABNF"></xref> with the EKT
        extensions to SDP Security Descriptions results in the following
        example SDP:</t>

        <figure>
          <artwork align="center"><![CDATA[
v=0
o=sam 2890844526 2890842807 IN IP4 192.0.2.5
s=SRTP Discussion
i=A discussion of Secure RTP
u=http://www.example.com/seminars/srtp.pdf
e=marge@example.com (Marge Simpson)
c=IN IP4 192.0.2.12
t=2873397496 2873404696
m=audio 49170 RTP/SAVP 0
a=crypto:1 AES_CM_128_HMAC_SHA1_80
  inline:WVNfX19zZW1jdGwgKCkgewkyMjA7fQp9CnVubGVz|2^20|1:4
  FEC_ORDER=FEC_SRTP EKT=AES_128|FE9C|AAE0
a=crypto:2 F8_128_HMAC_SHA1_80
  inline:MTIzNDU2Nzg5QUJDREUwMTIzNDU2Nzg5QUJjZGVm|2^20|1:4;
  inline:QUJjZGVmMTIzNDU2Nzg5QUJDREUwMTIzNDU2Nzg5|2^20|2:4
  FEC_ORDER=FEC_SRTP EKT=AES_128|FE9C|AAE0
]]></artwork>

          <postamble></postamble>
        </figure>

        <t>For legibility the SDP shows line breaks that are not present on
        the wire.</t>
      </section>
    </section>

    <section anchor="dtls-srtp-kt"
             title="Use of EKT with DTLS-SRTP Key Transport">
      <t>This document defines an extension to DTLS-SRTP called Key Transport.
      Using EKT with the DTLS-SRTP Key Transport extensions allows securely
      transporting SRTP keying material from one DTLS-SRTP peer to another, so
      the same SRTP keying material can be used by those peers and so those
      peers can process EKT keys. This combination of protocols is valuable
      because it combines the advantages of DTLS (strong authentication of the
      endpoint and flexibility) with the advantages of EKT (allowing secure
      multiparty RTP with loose coordination and efficient communication of
      per-source keys).</t>

      <section anchor="dtls-srtp-extensions"
               title="EKT Extensions to DTLS-SRTP">
        <t>This document adds a new TLS negotiated extension called "ekt".
        This adds a new TLS content type, EKT, and a new negotiated extension
        EKT. The negotiated extension MUST only be requested in conjunction
        with the "use_srtp" extension (Section 3.2 of <xref
        target="RFC5764"></xref>). The DTLS server indicates its support for
        EKT by including "dtls-srtp-ekt" in its SDP and "ekt" in its TLS
        ServerHello message. If a DTLS client includes "ekt" in its
        ClientHello, but does not receive "ekt" in the ServerHello, the DTLS
        client MUST NOT send DTLS packets with the "ekt" content-type.</t>

        <figure anchor="tls_datastructure"
                title="Additional TLS Data Structures">
          <preamble>Using the syntax described in <xref
          target="RFC6347">DTLS</xref>, the following
          structures are used:</preamble>

          <artwork align="center"><![CDATA[
enum {
  ekt_key(0),
  ekt_key_ack(1),
  ekt_key_error(254),
  (255)
} SRTPKeyTransportType;

struct {
  SRTPKeyTransportType keytrans_type;
  uint24 length;
  uint16 message_seq;
  uint24 fragment_offset;
  uint24 fragment_length;
  select (SRTPKeyTransportType) {
     case ekt_key:
        EKTkey;
   };
} KeyTransport;

enum {
 AES_128(0),
 AESKW_128(1),
 AESKW_192(2),
 AESKW_256(3),
} ektcipher;

struct {
  ektcipher EKT_Cipher;
  uint EKT_Key_Value<1..256>;
  uint EKT_Master_Salt<1..256>;
  uint16 EKT_SPI;
} EKTkey;
]]></artwork>
        </figure>

        <t>The diagram below shows a message flow of DTLS client and
        DTLS server using the DTLS-SRTP Key Transport extension. SRTP
        packets exchanged prior to the ekt_message are encrypted using
        the SRTP master key derived from the normal DTLS-SRTP key
        derivation function. After the ekt_key message, they can be
        encrypted using the EKT key.<list style="empty">
            <t>Editor's note: do we need reliability for the ekt_key
            messages?</t>
          </list></t>

        <figure anchor="tls_handshake_message_flow"
                title="Handshake Message Flow">
          <preamble></preamble>

          <artwork align="center"><![CDATA[
Client                                               Server

ClientHello + use_srtp + EKT
                            -------->
                              ServerHello + use_srtp + EKT
                                              Certificate*
                                        ServerKeyExchange*
                                       CertificateRequest*
                            <--------      ServerHelloDone
Certificate*
ClientKeyExchange
CertificateVerify*
[ChangeCipherSpec]
Finished                     -------->
                                        [ChangeCipherSpec]
                            <--------             Finished
SRTP packets                 <------->      SRTP packets
SRTP packets                 <------->      SRTP packets
ekt_key                      -------->
SRTP packets                 <------->      SRTP packets
SRTP packets                 <------->      SRTP packets
]]></artwork>
        </figure>

        <section title="Scaling to Large Groups">
          <t>In certain scenarios it is useful to perform DTLS-SRTP
          with a device that is not the RTP peer. A common scenario is
          multicast, where it is necessary to distribute the DTLS-SRTP
          (and EKT distribution) to several devices. To allow for
          this, a new SDP attribute, dtls-srtp-host, is defined which
          follows the general syntax specified in Section 5.13
          of <xref target="RFC4566"></xref>.  When signaled, it
          indicates this host controls the EKT keying for all group
          members.  For the dtls-srtp-host
          attribute: <list style="symbols">
              <t>the name is the ASCII string "dtls-srtp-host" (lowercase)</t>

              <t>the value is the IP address and port number used for
              DTLS-SRTP</t>

              <t>This is a media-level attribute and MUST NOT appear at the
              session level</t>
            </list>The formal description of the attribute is defined by the
          following <xref target="RFC5234">ABNF</xref> syntax:</t>

          <figure>
            <artwork><![CDATA[  
attribute = "a=dtls-srtp-host:" 
            dtls-srtp-host-info *(SP dtls-srtp-host-info)
host-info = nettype space addrtype space 
            connection-address space port CRLF]]></artwork>
          </figure>

          <t>Multiple IP/port pairs are provided for IPv6/IPv4 interworking,
          and to allow failover. The receiving host SHOULD attempt to use them
          in the order provided.</t>

          <figure>
            <preamble>An example of SDP containing the dtls-srtp-host
            attribute:</preamble>

            <artwork align="center"><![CDATA[         
v=0
o=sam 2890844526 2890842807 IN IP4 192.0.2.5
s=SRTP Discussion
i=A discussion of Secure RTP
u=http://www.example.com/seminars/srtp.pdf
e=marge@example.com (Marge Simpson)
c=IN IP4 192.0.2.12
t=2873397496 2873404696
m=audio 49170 UDP/TLS/RTP/SAVP 0
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=dtls-srtp-ekt
a=dtls-srtp-host:IN IP4 192.0.2.13 56789
]]></artwork>

            <postamble>For legibility the SDP shows line breaks that are not
            present on the wire.</postamble>
          </figure>
        </section>
      </section>

      <section title="Offer/Answer Considerations">
        <t>This section describes Offer/Answer considerations for the use of
        EKT together with DTLS-SRTP for unicast and multicast streams. The
        offerer and answerer MUST follow the procedures specified in <xref
        target="RFC5764"></xref> as well as the following ones.</t>

        <t>As most DTLS-SRTP processing is performed on the media channel,
        rather than in SDP, there is little processing performed in SDP other
        than informational and to redirect DTLS-SRTP to an alternate host.
        Advertising support for the extension is necessary in SDP because in
        some cases it is required to establish an SRTP call. For example, a
        mixer may be able to only support SRTP listeners if those listeners
        implement DTLS Key Transport (because it lacks the CPU cycles
        necessary to encrypt SRTP uniquely for each listener).</t>

        <section title="Generating the Initial Offer">
          <t>The initial offer contains a new SDP attribute, "dtls-srtp-ekt",
          which contains no value. This indicates the offerer is capable of
          supporting DTLS-SRTP with EKT extensions, and indicates the desire
          to use the "ekt" extension during the DTLS-SRTP handshake. If the
          offerer wants another host to perform DTLS-SRTP-EKT processing, it
          also includes the dtls-srtp-host attribute in its offer (<xref
          target="dtls-srtp-extensions"></xref>).</t>

          <figure>
            <preamble>An example of SDP containing the dtls-srtp-ekt
            attribute::</preamble>

            <artwork align="center"><![CDATA[         
v=0
o=sam 2890844526 2890842807 IN IP4 192.0.2.5
s=SRTP Discussion
i=A discussion of Secure RTP
u=http://www.example.com/seminars/srtp.pdf
e=marge@example.com (Marge Simpson)
c=IN IP4 192.0.2.12
t=2873397496 2873404696
m=audio 49170 UDP/TLS/RTP/SAVP 0
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=dtls-srtp-ekt
]]></artwork>

            <postamble>For legibility the SDP shows line breaks that are not
            present on the wire.</postamble>
          </figure>

          <t></t>
        </section>

        <section title="Generating the Initial Answer">
          <t>Upon receiving the initial offer, the presence of the
          dtls-srtp-ekt attribute indicates a desire to receive the EKT
          extension in the DTLS-SRTP handshake. The presence of the
          dtls-srtp-host attribute indicates an alternate host to send the
          DTLS-SRTP handshake (instead of the host on the c/m lines). DTLS
          messages should be constructed according to those two
          attributes.</t>

          <t>The SDP answer SHOULD contain the dtls-srtp-ekt attribute to
          indicate the answerer understands dtls-srtp. It should only contain
          the dtls-srtp-host attribute if the answerer also wishes to offload
          its DTLS-SRTP processing to another host.</t>
        </section>

        <section title="Processing the Initial Answer">
          <t>The presence of the dtls-srtp-ekt attribute indicates a desire by
          the answerer to perform DTLS-SRTP with EKT extensions, and the
          dtls-srtp-host attribute indicates an alternate host for DTLS-SRTP
          processing.</t>

          <t>After successful negotiation of the key_transport extension, the
          DTLS client and server MAY exchange SRTP packets, encrypted using
          the KDF described in <xref target="RFC5764"></xref>. This is normal
          and expected, even if Key Transport was negotiated by both sides, as
          neither side may (yet) have a need to alter the SRTP key. However,
          it is also possible that one (or both) peers will immediately send
          new_srtp_key message before sending any SRTP, and also possible that
          SRTP, encrypted with an unknown key, may be received before the
          new_srtp_key message is received.</t>
        </section>

        <section title="Modifying the Session">
          <t>As DTLS-SRTP-EKT processing is done on the DTLS-SRTP channel
          (media channel) rather than signaling, no special processing for
          modifying the session is necessary.</t>
        </section>
      </section>
    </section>

    <section anchor="mikey" title="Use of EKT with MIKEY">
      <t>The advantages outlined in Section 1 are useful in some scenarios in
      which MIKEY is used to establish SRTP sessions. In this section, we
      briefly review MIKEY and related work, and discuss these scenarios.</t>

      <t>An SRTP sender or a group controller can use MIKEY to establish a
      SRTP cryptographic context. This capability includes the distribution of
      a TEK generation key (TGK) or the TEK itself, security policy payload,
      crypto session bundle ID (CSB_ID) and a crypto session ID (CS_ID). The
      TEK directly maps to an SRTP master key, whereas the TGK is used along
      with the CSB_ID and a CS_ID to generate a TEK. The CS_ID is used to
      generate multiple TEKs (SRTP master keys) from a single TGK. For a media
      stream in SDP, MIKEY allocates two consecutive numbers for the crypto
      session IDs, so that each direction uses a different SRTP master key
      (see <xref target="RFC4567"></xref>).</t>

      <t>The MIKEY specification <xref target="RFC3830"></xref> defines three
      modes to exchange keys, associated parameters and to protect the MIKEY
      message: pre-shared key, public-key encryption and Diffie-Hellman key
      exchange. In the first two modes the MIKEY initiator only chooses and
      distributes the TGK or TEK, whereas in the third mode both MIKEY
      entities (the initiator and responder) contribute to the keys. All three
      MIKEY modes have in common that for establishing a SRTP session the
      exchanged key is valid for the send and receive direction. Especially
      for group communications it is desirable to update the SRTP master key
      individually per direction. EKT provides this property by distributing
      the SRTP master key within the SRTP/SRTCP packet.</t>

      <t>MIKEY already supports synchronization of ROC values between the
      MIKEY initiator and responder. The SSRC / ROC value pair is part of the
      MIKEY Common Header payload. This allows providing the current ROC value
      to late joiners of a session. However, in some scenarios a key
      management based ROC synchronization is not sufficient. For example, in
      mobile and wireless environments, members may go in and out of coverage
      and may miss a sequence number overrun. In point-to-multipoint
      translator scenarios it is desirable to not require the group controller
      to track the ROC values of each member, but to provide the ROC value by
      the originator of the SRTP packet. A better alternative to synchronize
      the ROC values is to send them directly via SRTP/SRTCP, as EKT does. A
      separate SRTP extension is being proposed <xref target="RFC4771"></xref>
      to include the ROC as part of a modified authentication tag. Unlike EKT,
      this extension uses only SRTP and not SRTCP as its transport and does
      not allow updating the SRTP master key.</t>

      <t>Besides the ROC, MIKEY synchronizes also the SSRC values of the SRTP
      streams. Each sender of a stream sends the associated SSRC within the
      MIKEY message to the other party. If a SRTP session participant starts a
      new SRTP source or a new participant is added to a group, subsequent SDP
      offer/answer and MIKEY exchanges are necessary to update the SSRC
      values. EKT improves these scenarios by updating the keys and SSRC
      values without coordination on the signaling channel. With EKT, SRTP can
      handle early media, since the EKT SPI allows the receiver to identify
      the cryptographic keys and parameters used by the source.</t>

      <t>The MIKEY specification <xref target="RFC3830"></xref> suggests the
      use of unicast for rekeying. This method does not scale well to large
      groups or interactive groups. The EKT extension of SRTP/SRTCP provides a
      solution for rekeying the SRTP master key and for ROC/SSRC
      synchronization. EKT is not a substitution for MIKEY, but rather a
      complementary addition to address the above described limitations of
      MIKEY.</t>

      <t>In the next section we provide an extension to MIKEY for support of
      EKT. EKT can be used only with the pre-shared key or public-key
      encryption MIKEY mode of <xref target="RFC3830"></xref>. The
      Diffie-Hellman exchange mode is not suitable in conjunction with EKT,
      because it is not possible to establish one common EKT key over multiple
      EKT entities. Additional MIKEY modes specified in separate documents are
      not considered for EKT.</t>

      <section title="EKT extensions to MIKEY">
        <t>In order to use EKT with MIKEY, the EKT cipher, EKT key and EKT SPI
        must be negotiated in the MIKEY message exchange.</t>

        <t>For EKT we specify a new SRTP Policy Type in the Security Policy
        (SP) payload of MIKEY (see Section 6.10 of <xref
        target="RFC3830"></xref>). The SP payload contains a set of policies.
        Each policy consists of a number Policy Param TLVs.</t>

        <figure anchor="EKT_SecPol" title="EKT Security Policy">
          <preamble></preamble>

          <artwork align="center"><![CDATA[
Prot type  | Value
-------------------
EKT        | TBD (will be requested from IANA)
]]></artwork>

          <postamble>For legibility the SDP shows line breaks that are not
          present on the wire.</postamble>
        </figure>

        <t>The EKT Security Policy has one parameter representing the EKT
        cipher.</t>

        <figure anchor="EKT_SecPolParam"
                title="EKT Security Policy Parameters">
          <preamble></preamble>

          <artwork align="center"><![CDATA[
Type | Meaning                     | Possible values
----------------------------------------------------
   0 | EKT cipher                  | see below
]]></artwork>
        </figure>

        <figure anchor="EKT_CipherParam" title="EKT Cipher Parameters">
          <preamble></preamble>

          <artwork align="center"><![CDATA[
EKT cipher | Value
-------------------
AES_128    |  0
AESKW_128  |  1
AESKW_192  |  2
AESKW_256  |  3
]]></artwork>

          <postamble></postamble>
        </figure>

        <t>AES_128 is the default value for the EKT cipher.</t>

        <t>The two mandatory EKT parameters (EKT_Key and EKT_SPI) are
        transported in the MIKEY KEMAC payload within one separate Key Data
        sub-payload. As specified in Section 6.2 of <xref
        target="RFC3830"></xref>, the KEMAC payload carries the TEK Generation
        Key (TGK) or the Traffic Encryption Key (TEK). One or more TGKs or
        TEKs are carried in individual Key Data sub-payloads within the KEMAC
        payload. The KEMAC payload is encrypted as part of MIKEY. The Key Data
        sub- payload, specified in Section 6.13 of <xref
        target="RFC3830"></xref>, has the following format:</t>

        <figure anchor="Key_Data_subpayload"
                title="Key Data Sub-Payload of MIKEY">
          <preamble></preamble>

          <artwork align="center"><![CDATA[
                     1                   2                   3
 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Payload  | Type  |  KV   | Key data length               |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
:                         Key data                              :
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
: Salt length (optional)        ! Salt data (optional)          :
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
:                        KV data (optional)                     :
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
]]></artwork>

          <postamble></postamble>
        </figure>

        <t>These fields are described below:<list style="hanging">
            <t hangText="Type:">4 bits in length, indicates the type of key
            included in the payload. We define Type = TBD (will be requested
            from IANA) to indicate transport of the EKT key.</t>

            <t hangText="KV:">(4 bits): indicates the type of key validity
            period specified. KV=1 is currently specified as an SPI. We use
            that value to indicate the KV_data contains the ETK_SPI for the
            key type EKT_Key.  KV_data would be 16 bits in length, but it is
            also possible to interpret the length from the 'Key data len'
            field. KV data MUST NOT be optional for the key type EKT_Key when
            KV = 1.</t>

            <t hangText="Salt length, Salt Data:">These optional fields SHOULD
            be omitted for the key type EKT_Key, if the SRTP master salt is
            already present in the TGK or TEK Key Data sub-payload. The
            EKT_Key sub-payload MUST contain a SRTP master salt, if the SRTP
            master salt is not already present in the TGK or TEK Key Data
            sub-payload.</t>

            <t hangText="KV Data:">length determined by Key Data Length
            field.</t>
          </list></t>
      </section>

      <section title="Offer/Answer considerations ">
        <t>This section describes Offer/Answer considerations for the use of
        EKT together with MIKEY for unicast streams. The offerer and answerer
        MUST follow the procedures specified in <xref target="RFC3830"></xref>
        and <xref target="RFC4567"></xref> as well as the following ones.</t>

        <section title="Generating the Initial Offer">
          <t>If it is intended to use MIKEY together with EKT, the offerer
          MUST include at least one MIKEY key-mgmt attribute with one EKT_Key
          Key Data sub-payload and the EKT_Cipher Security Policy payload.
          MIKEY can be used on session or media level. On session level, MIKEY
          provides the keys for multiple SRTP sessions in the SDP offer. The
          EKT SPI references a EKT parameter set including the Secure RTP
          parameters as specified in Section 8.2 in <xref
          target="RFC3711"></xref>. If MIKEY is used on session level, it is
          only possible to use one EKT SPI value. Therefore, the session-level
          MIKEY message MUST contain one SRTP Security Policy payload only,
          which is valid for all related SRTP media lines. If MIKEY is used on
          media level, different SRTP Security Policy parameters (and
          consequently different EKT SPI values) can be used for each media
          line. If MIKEY is used on session and media level, the medial level
          content overrides the session level content.</t>

          <t>EKT requires a single shared SRTP master salt between all
          participants in the direct SRTP session. If a MIKEY key-mgmt
          attribute contains more than one TGK or TEK Key Data sub-payload,
          all the sub-payloads MUST contain the same master salt value.
          Consequently, the EKT_Key Key Data sub-payload MAY also contain the
          same salt or MAY omit the salt value. If the SRTP master salt is not
          present in the TGK and TEK Key Data sub-payloads, the EKT_Key
          sub-payload MUST contain a master salt.</t>
        </section>

        <section title="Generating the Initial Answer">
          <t>For each media line in the offer using MIKEY, provided on session
          or/and on media level, the answerer examines the related MIKEY
          key-mgmt attributes for the presence of EKT parameters. In order to
          accept the offered key-mgmt attribute, the MIKEY message MUST
          contain one EKT_Key Key Data sub-payload and the EKT_Cipher Security
          Policy payload. The answerer examines also the existence of a SRTP
          master salt in the TGK/TEK and/or the EKT_Key sub-payloads. If
          multiple salts are available, all values MUST be equal. If the salt
          values differ or no salt is present, the key-mgmt attribute MUST be
          considered as invalid.</t>

          <t>The MIKEY responder message in the SDP answer does not contain a
          MIKEY KEMAC or Security Policy payload and consequently does not
          contain any EKT parameters. If the key-mgmt attribute for a media
          line was accepted by the answerer, the EKT parameter set of the
          offerer is valid for both directions of the SRTP session.</t>
        </section>

        <section title="Processing the Initial Answer">
          <t>On reception of the answer, the offerer examines if EKT has been
          accepted for the offered media lines. If a MIKEY key-mgmt attribute
          is received containing a valid MIKEY responder message, EKT has been
          successfully negotiated. On receipt of a MIKEY error message, EKT
          negotiation has failed. For example, this may happen if an EKT
          extended MIKEY initiator message is sent to a MIKEY entity not
          supporting EKT. A MIKEY error code 'Invalid SP' or 'Invalid DT' is
          returned to indicate that the EKT_Cipher Security Policy payload or
          the EKT_Key sub-payload is not supported. In this case, the offerer
          may send a second SDP offer with a MIKEY key-mgmt attribute without
          the additional EKT extensions.</t>

          <t>This behavior can be improved by defining an additional key-mgmt
          prtcl-id value 'mikeyekt' and offering two key-mgmt SDP attributes.
          One attribute offers MIKEY together with EKT and the other one
          offers MIKEY without EKT. This is for further discussion.</t>
        </section>

        <section title="Modifying the Session">
          <t>Once a SRTP stream has been established, a new offer/answer
          exchange can modify the session including the EKT parameters. If the
          EKT key or EKT cipher is modified (i.e., a new EKT parameter set is
          created) the offerer MUST also provide a new EKT SPI value. The
          offerer MUST NOT remap an existing EKT SPI value to a new EKT
          parameter set. Similar, a modification of the SRTP Security Policy
          leads to a new EKT parameter set and requires a fresh EKT SPI, even
          the EKT key or cipher did not change.</t>

          <t>Once EKT is being used, the SRTP master salt for the SRTP session
          MUST NOT be changed. The salt in the Key Data sub-payloads within
          the subsequent offers MUST be the same as the one already used.</t>
<!-- dwing comment
Do we want to recommend any action if they differ?  Perhaps restart EKT,
or abort the entire SRTP session immediately?
-->


          <t>After EKT has been successfully negotiated for a session and a
          SRTP master key has been transported by EKT, it is difficult to
          switch back to a pure MIKEY based key exchange in a synchronized
          way. Therefore, once EKT is being used for a session, EKT MUST be
          used also in all subsequent offer/answer exchanges for that
          session.</t>
        </section>
      </section>
    </section>

    <section title="Using EKT for interoperability between key management systems" anchor="interop">
      <t>
	
	A media gateway (MGW) can provide interoperability between an
	SRTP-EKT endpoint and a non-EKT SRTP endpoint.  When doing
	this function, the MGW can perform non-cryptographic
	transformations on SRTP packets outlined above.  However, there are
	some uses of cryptography that will be required for that
	gateway.
	
	If a new SRTP master key is communicated to the MGW (via EKT from the
	EKT leg, or via Security Descriptions from the Security Descriptions
	leg), the MGW needs to convert that information for the other leg, and
	that process will incur some cryptographic operations.  Specifically,
	if the new key arrived via EKT, that must be decrypted and then sent
	in Security Descriptions; likewise, if a new key arrives via Security
	Descriptions that must be encrypted via EKT and sent in SRTP/SRTCP.

      </t>
      <t>Additional non-normative information can be found in <xref 
target="appendix_interworking"></xref>.</t>

    </section>

    <section anchor="rationale" title="Design Rationale">
      <t>From <xref target="RFC3550"></xref>, a primary function of RTCP is to
      carry the CNAME, a "persistent transport-level identifier for an RTP
      source" since "receivers require the CNAME to keep track of each
      participant." EKT works in much the same way, using SRTCP to carry
      information needed for the proper processing of the SRTP traffic.</t>

      <t>With EKT, SRTP gains the ability to synchronize the creation of
      cryptographic contexts across all of the participants in a single
      session. This feature provides some, but not all, of the functionality
      that is present in IKE phase two (but not phase one). Importantly, EKT
      does not provide a way to indicate SRTP options.</t>

      <t>With EKT, external signaling mechanisms provide the SRTP options and
      the EKT Key, but need not provide the key(s) for each individual SRTP
      source. EKT provides a separation between the signaling mechanisms and
      the details of SRTP. The signaling system need not coordinate all SRTP
      streams, nor predict in advance how many streams will be present, nor
      communicate SRTP-level information (e.g., rollover counters) of current
      sessions.</t>

      <t>EKT is especially useful for multi-party sessions, and for the case
      where multiple RTP sessions are sent to the same destination transport
      address (see the example in the definition of "RTP session" in <xref
      target="RFC3550"></xref>). A SIP offer that is forked in parallel (sent
      to multiple endpoints at the same time) can cause multiple RTP sessions
      to be sent to the same transport address, making EKT useful for use with
      SIP.</t>

      <t>EKT can also be used in conjunction with a scalable group-key
      management system like <xref target="RFC6407">GDOI</xref>. Such a system
      provides a secure entity authentication method and a way to revoke group
      membership, both of which are out of scope of EKT.</t>

      <!--
There are several scenarios in which there are more than two
participants in an SRTP sessions, besides the important case of
multi-party conferences.  For example, a session may be recorded (by a
voicemail system) or monitored (by a supervisor), or may
unintentionally include a third party (for example, a receptionist)
due to a race condition in call signaling.
A session can start out as conversational voice, then switch codecs
and SSRC values to send a fax transmission.
-->

      <t>It is natural to use SRTCP to transport encrypted keying material for
      SRTP, as it provides a secure control channel for (S)RTP. However, there
      are several different places in SRTCP in which the encrypted SRTP master
      key and ROC could be conveyed. We briefly review some of the
      alternatives in order to motivate the particular choice used in this
      specification. One alternative is to have those values carried as a new
      SDESC item or RTCP packet. This would require that the normal SRTCP
      encryption be turned off for the packets containing that SDESC item,
      since on the receiver's side, SRTCP processing completes before the RTCP
      processing starts. This tension between encryption and the desire for
      RTCP privacy is highly undesirable. Additionally, this alternative makes
      SRTCP dependent upon the parsing of the RTCP compound packet, which adds
      complexity. It is simpler to carry the encrypted key in a new SRTCP
      field. One way to do this and to be backwards compatible with the
      existing specification is to define a new crypto function that
      incorporates the encrypted key. We define a new authentication transform
      because EKT relies on the normal SRTCP authentication to provide
      implicit authentication of the encrypted key.<!--
Recall that SRTP allows a transform definition to "add steps to
the packet processing, or even add fields to the SRTP/SRTCP packets."
--></t>

      <t>An SRTP packet containing an SSRC that has not been seen will be
      discarded. This practice may induce a burst of packet loss at the outset
      of an SRTP stream, due to the loss or reorder of the first SRTCP packet
      with the EKT containing the key and rollover counter for that stream.
      However, this practice matches the conservative RTP memory-allocation
      strategy; many existing applications accept this risk of initial packet
      loss. Alternatively, implementations may wish to delay discarding such
      packets for a short period of time as described in <xref
      target="SynchronizingOperation"></xref>.</t>

<!-- dwing comment
dwing-comment-length:  the two paragraphs below seems to conflict with the 
lengths mentioned earlier in the document (search for the string 
dwing-comment-length)


      <t>When EKT is carried in SRTCP, it adds eight additional bytes to each
      SRTCP packet, plus the length of the EKT Ciphertext field. Using
      the SRTP and EKT defaults, the total overhead is 24 bytes. This overhead
      does not detract from the total bandwidth used by SRTP, since it is
      included in the RTCP bandwidth computation. However, it will cause the
      control protocol to issue packets less frequently.</t>

-->

      <t>The main motivation for the use of the variable-length format is
      bandwidth conservation. If EKT is used of SRTP, there will be a loss of
      bandwidth due to the additional 24 bytes in each RTP packet. For some
      applications, this bandwidth loss is significant.<!--
A data format
that uses this approach is defined in <xref target="appendix"/>.  We
leave this point open for discussion.
--></t>

      <!--
     <t>
       The main motivation for defining the ability to run EKT over
       SRTP instead of RTCP are the unfortunate facts that RTCP is not
       always available, and that
       some firewalls and NAT devices can pass RTP but not RTCP.
     </t>
-->

      <section title="Alternatives">
	<t>
	  In its current design, EKT requires that the Master Salt be
	  established out of band.  That requirement is undesirable.
	  In an offer/answer environment, it forces the answerer to
	  re-use the same Master Salt value used by the offerer.  The
	  Master Salt value could be carried in EKT packets though
	  that would consume yet more bandwidth.
	</t>
	<t>
	  In some scenarios, two SRTP sessions may be combined into a
	  single session.  When using EKT in such sessions, it is
	  desirable to have an SPI value that is larger than 15 bits,
	  so that collisions between SPI values in use in the two
	  different sessions are unlikely (since each collision would
	  confuse the members of one of the sessions.)
	  </t>
	<t>
	  An alternative that addresses both of these needs is as
	  follows: the SPI value can be lengthed from 15 bits to 63
	  bits, and the Master Salt can be identical to, or
	  constructed from, the SPI value.  SRTP conventionally uses a
	  14-byte Master Salt, but shorter values are acceptable.
	  This alternative would add six bytes to each EKT packet;
	  that overhead may be a reasonable tradeoff for addressing
	  the problems outlined above.  
	  </t>
      </section>

    </section>

    <section anchor="sec" title="Security Considerations">
      <t>With EKT, each SRTP sender and receiver can generate distinct SRTP
      master keys. This property avoids any security concern over the re-use
      of keys, by empowering the SRTP layer to create keys on demand. Note
      that the inputs of EKT are the same as for SRTP with key-sharing: a
      single key is provided to protect an entire SRTP session. However, EKT
      provides complete security, even in the absence of further out-of-band
      coordination of SSRCs, and even when SSRC values collide.</t>

      <t>In order to avoid potential security issues, the SRTP authentication
      tag length used by the base authentication method MUST be at least ten
      octets.</t>
<!-- dwing comment
ten octets exceeds requirement of Suite B Profile for DTLS-SRTP, which
only needs 8 octets.
-->

      <t>
	The presence of the SSRC in the EKT_Plaintext ensures that an
	attacker cannot substitute an EKT_Ciphertext from one SRTP
	stream into another SRTP stream, even if those two streams are
	using the same SRTP master key.  This is important because
	some applications may use the same master key for multiple
	streams.
      </t>	
      <t>
	An attacker who strips a Full_EKT_Field from an SRTP packet may prevent
	the intended receiver of that packet from being able to decrypt it.
	This is a minor denail of service vulnerability.   Similarly, an
	attacker who adds a Full_EKT_Field can disrupt service.  
      </t>

<!-- dwing commment
Following paragraph in security considerations added by dwing
-->

<t>An attacker could send packets containing either Short EKT Tag or
Full EKT Tag, in an attempt to consume additional CPU resources of the
receiving system.  In the case of the Short EKT Tag, this field is
stripped and normal SRTP or SRTCP processing is performed.  In the
case of the Full EKT Tag, the attacker would have to have guessed or
otherwise determined the SPI being used by the receiving system.  If
an invalid SPI is provided by the attacker, processing stops.  If a
valid SPI is provided by the attacker, the receiving system will
decrypt the EKT ciphertext and return an authentication failure (Step
3 of <xref target="inbound"></xref>).</t>



      <t>
	An attacker learns from EKT when SRTP Master Keys change.
      </t>
      <t>
	The EKT Cipher MUST be at least as strong as the encryption and
	authentication operations used in SRTP.  
      </t>
      <t>
	Part of the EKT_Plaintext is known, or easily guessable to an
	attacker.  Thus, the EKT Cipher MUST resist known plaintext attacks.
	In practice, this requirement does not impose any restrictions 
	on our choices, since the ciphers in use provide high security
	even when much plaintext is known.
      </t>
        <t>An EKT cipher MUST resist attacks in which both ciphertexts and
        plaintexts can be adaptively chosen. For each randomly chosen key, the
        encryption and decryption functions cannot be distinguished from a
        random permutation and its inverse with non-negligible advantage. This
        must be true even for adversaries that can query both the encryption
        and decryption functions adaptively. The advantage is defined as the
        difference between the probability that the adversary will identify
        the cipher as such and the probability that the adversary will
        identify the random permutation as the cipher, when each case is
        equally likely.</t>
    </section>

    <section anchor="iana" title="IANA Considerations">
      <t>IANA is requested to register EKT into
the <xref target="iana-sdp-sdesc">SRTP Session Parameter
registry</xref>.</t>

<t>IANA is requested to register dtls-srtp-ekt and dtls-srtp-host into
the att-field table of the <xref target="iana-sdp-attr">SDP Attributes
registry</xref>.</t>


      <!--MIKEY Registrations-->

      <t>We request the following IANA assignments from existing MIKEY IANA
      tables: <list style="symbols">
          <t>From the Key Data payload name spaces, a value to indicate the
          type as the 'EKT_Key'.</t>

          <t>From the Security Policy table name space, a new value to be
          assigned for 'EKT' (see <xref target="EKT_SecPol"></xref>).</t>
        </list></t>

      <t>Furthermore, we need the following two new IANA registries created,
      populated with the initial values in this document. New values for both
      of these registries can be defined via <xref
      target="RFC5226">Specification Required</xref>. <list style="symbols">
          <t>EKT parameter type (initially populated with the list from <xref
          target="EKT_SecPolParam"></xref>)</t>

          <t>EKT cipher (initially populated with the list from <xref
          target="EKT_CipherParam"></xref>)</t>
        </list></t>
    </section>

    <section title="Acknowledgements">
      <t>Thanks to Lakshminath Dondeti for assistance with earlier versions of
      this document. Thanks to Nermeen Ismail, Eddy Lem, and Rob Raymond for
      fruitful discussions and comments. Thanks to Romain Biehlmann for his
      encouragement to add support DTLS-SRTP-EKT key servers for multicast.
      Thanks to Felix Wyss for his review and comments regarding ciphers.</t>
    </section>
  </middle>

  <back>
    <references title="Normative References">
      <reference anchor="FIPS197">
        <front>
          <title>The Advanced Encryption Standard (AES)</title>

          <author fullname="National Institute of Standards and Technology   (NIST)">
            <organization />
          </author>
        </front>

        <seriesInfo name="FIPS-197"
                    value="Federal Information Processing Standard" />
      </reference>

      <!--
	<reference anchor="LRW02">
         <front>
           <title>Tweakable Block Ciphers</title>
	  <author initials="M." surname="Liskov" fullname="Moses Liskov">
	      <organization/>
           </author>
	  <author initials="R." surname="Rivest" fullname="Ron Rivest">
	    <organization/>
	  </author>
	  <author initials="D." surname="Wagner" fullname="David Wagner">
	    <organization/>
	  </author>
	  <date month="August" year="2002"/>
         </front>
         <seriesInfo name="CRYPTO 2002" value="Springer-Verlag" />
         </reference>



	<reference anchor="SDES">
	<front>
	  <title> Session Description Protocol Security Descriptions
             for Media Streams </title>
	  <author initials="F." surname="Andreasen"
		  fullname = "Flemming Andreasen">
	  <organization/>
	  </author>
	  <author initials="M." surname="Baugher"
		  fullname = "Mark Baugher">
	  <organization/>
	  </author>
	  <author initials="D." surname="Wing"
		  fullname = "Dan Wing">
	  <organization/>
	  </author>
	</front>
	<seriesInfo name="Work In Progress." value="<draft-ietf-mmusic-sdescriptions-11.txt>"/>
	</reference>

-->

      <!--

	<reference anchor="RCC">
	<front>
	  <title> Integrity Transform Carrying Roll-over Counter </title>
	  <author initials="V." surname="Lehtovirta"
		  fullname = "Vesa Lehtovirta">
	  <organization/>
	  </author>
	  <author initials="M." surname="Naslund"
		  fullname = "Mats Naslund">
	  <organization/>
	  </author>
	  <author initials="K." surname="Norrman"
		  fullname = "Karl Norrman">
	  <organization/>
	  </author>
	</front>
	<seriesInfo name="Work In Progress." value="<draft-lehtovirta-srtp-rcc-01.txt>"/>
	</reference>

	<reference anchor="KEYID">
	<front>
	  <title> The Key ID Information Type for the General
	  Extension Payload in MIKEY
	  </title>
	  <author initials="E." surname="Carrara"
		  fullname = "Elisabetta Carrara"> <organization/>
	  </author>
	  <author initials="V." surname="Lehtovirta"
		  fullname = "Vesa Lehtovirta"> <organization/>
	  </author>
	  <author initials="K." surname="Norrman"
		  fullname = "Karl Norrman">
	  <organization/>
	  </author>
	</front>
	<seriesInfo name="Work In Progress." value="<draft-ietf-msec-newtype-keyid-04.txt>"/>
	</reference>

-->

      &rfc4563;

      &rfc4567;

      &rfc4568;

      &rfc4771;

      &rfc2119;

      &rfc3261;

      &rfc3264;



      &rfc3550;

      &rfc4648;

      &rfc3711;

      &rfc5234;

 &rfc6347; 

      &rfc5764;

      &rfc4566;
      &rfc5226;
    </references>

    <references title="Informative References">
      &rfc3830;
      &rfc5649;

      &rfc4301;
      &rfc6407;

      <reference anchor="iana-sdp-attr"
                 target="http://www.iana.org/assignments/sdp-parameters/sdp-parameters.xml">
        <front>
          <title>SDP Parameters</title>
          <author fullname="IANA" surname="IANA">
            <organization></organization>
          </author>
          <date year="2011" />
        </front>
      </reference>

      <reference anchor="iana-sdp-sdesc"
                 target="http://www.iana.org/assignments/sdp-security-descriptions/sdp-security-descriptions.xml">
        <front>
          <title>SDP Security Descriptions</title>
          <author fullname="IANA" surname="IANA">
            <organization></organization>
          </author>
          <date year="2011" />
        </front>
      </reference>


    </references>

    <!--
 <section anchor="appendix" title="Alternate Format">
<t>
In this appendix we describe an alternate Authentication Tag format,
which is intended for the purposes of discussion.  It allows a sender
to optionally omit the EKT data from a packet.  As discussed in
<xref target="rationale"/>, this feature is desirable because it
allows bandwidth conservation, and it gives the sender the discretion
of when to send the EKT data, without any need for external
coordination.
</t>
<t>
The alternate format is shown in <xref target="altFormat"/>.  The top
diagram shows the format in the case that the final bit is set to one,
in which case all of the EKT fields are present.  The bottom diagram
shows the format in the case that the final bit is set to zero, in
which case the EKT Ciphertext, Rollover Counter, Initial
Sequence Number, and Security Parameter Index fields are absent.  The
Reserved field MUST be set to the all-zero value.  These two cases can
always be unambiguously distinguished by the final bit, or by checking
to see if the final byte in the packet has the all-zero value.
</t>
<figure anchor="altFormat" title="Alternate EKT Field format.">
<artwork>
     0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     :                   Base Authentication Tag                     :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     :                     EKT Ciphertext                      :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |                       Rollover Counter                        |
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |    Initial Sequence Number    |   Security Parameter Index  |1|
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

     0                   1                   2                   3
      0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     :                   Base Authentication Tag                     :
     +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
     |  Reserved   |0|
     +-+-+-+-+-+-+-+-+
</artwork>
</figure>
<t> In the alternate format, the SPI field is 15 bits long, instead of
16 bits long.  Sender-side implementations using the existing format
can achieve interoperability with the alternate format by selecting
SPI values have a final bit that is equal to one.  Receiver
implementations using the existing format can interoperate with the
alternate format if SPI values ending in one are used, and the sender
always sends all of the EKT fields.
</t>
<t>
The main motivation for the alternate format is the case when RTCP is
not available and EKT data is carried by RTP, and bandwidth
conservation is important.  However, it may be acceptable to use it
for RTCP as well.
</t>

</section>

-->

    <section anchor="appendix_interworking" title="Using EKT to Optimize Interworking DTLS-SRTP with Security Descriptions">
      <t>Today, <xref target="RFC4568">SDP Security Descriptions</xref> is
      used for distributing SRTP keys in several different IP PBX systems and
      is expected to be used by 3GPP's Long Term Evolution (LTE). The IP PBX
      systems are typically used within a single enterprise, and LTE is used
      within the confines of a mobile operator's network. A Session Border
      Controller is a reasonable solution to interwork between Security
      Descriptions in one network and DTLS-SRTP in another network. For
      example, a mobile operator (or an Enterprise) could operate Security
      Descriptions within their network and DTLS-SRTP towards the
      Internet.</t>

      <t>However, due to the way Security Descriptions and DTLS-SRTP manage
      their SRTP keys, such an SBC has to authenticate, decrypt, re-encrypt,
      and re-authenticate the SRTP (and SRTCP) packets in one direction, as
      shown in <xref target="interworking-expensive"></xref>, below. This is
      computationally expensive.</t>

      <figure anchor="interworking-expensive"
              title="Interworking Security Descriptions and DTLS-SRTP">
        <preamble></preamble>

        <artwork align="center"><![CDATA[
RFC4568 endpoint            SBC            DTLS-SRTP endpoint
      |                      |                       |
 1.   |---key=A------------->|                       |
 2.   |                      |<-DTLS-SRTP handshake->|
 3.   |<--key=B--------------|                       |
 4.   |                      |<--SRTP, encrypted w/B-|
 5.   |<-SRTP, encrypted w/B-|                       |
 6.   |-SRTP, encrypted w/A->|                       |
 7.   |            (decrypt, re-encrypt)             |
 8.   |                      |-SRTP, encrypted w/C-->|
      |                      |                       |]]></artwork>

        <postamble></postamble>
      </figure>

      <t>The message flow is as follows (similar steps occur with SRTCP):<list
          style="numbers">
          <t>The <xref target="RFC4568">Security Descriptions</xref> endpoint
          discloses its SRTP key to the SBC, using a=crypto in its SDP.</t>

          <t>SBC completes DTLS-SRTP handshake. From this handshake, the SBC
          derives the SRTP key for traffic from the DTLS-SRTP endpoint (key B)
          and to the DTLS-SRTP endpoint (key C).</t>

          <t>The SBC communicates the SRTP encryption key (key B) to the
          Security Descriptions endpoint (using a=crypto). (There is no way,
          with DTLS-SRTP, to communicate the Security Descriptions key to the
          DTLS-SRTP key endpoint.)</t>

          <t>The DTLS-SRTP endpoint sends an SRTP key, encrypted with its key
          B. This is received by the SBC.</t>

          <t>The received SRTP packet is simply forwarded; the SBC does not
          need to do anything with this packet as its key (key B) was already
          communicated in step 3.</t>

          <t>The Security Descriptions endpoint sends an SRTP packet,
          encrypted with its key A.</t>

          <t>The SBC has to authenticate and decrypt the SRTP packet (using
          key A), and re-encrypt it and generate an HMAC (using key C).</t>

          <t>The SBC sends the new SRTP packet.</t>
        </list></t>

      <t>If EKT is deployed on the DTLS-SRTP endpoints, EKT helps to avoid the
      computationally expensive operation so the SBC does not need not perform
      any per-packet operations on the SRTP (or SRTCP) packets in either
      direction. With EKT the SBC can simply forward the SRTP (and SRTCP)
      packets in both directions without per-packet HMAC or cryptographic
      operations.</t>

      <t>To accomplish this interworking, DTLS-SRTP EKT must be supported on
      the DTLS-SRTP endpoint, which allows the SBC to transport the Security
      Description key to the EKT endpoint and send the DTLS-SRTP key to the
      Security Descriptions endpoint. This works equally well for both
      incoming and outgoing calls. An abbreviated message flow is shown in
      <xref target="interworking-cheap"></xref>, below.</t>

      <figure anchor="interworking-cheap"
              title="Interworking Security Descriptions and EKT">
        <preamble></preamble>

        <artwork align="center"><![CDATA[
RFC4568 endpoint            SBC          DTLS-SRTP endpoint
      |                      |                       |
 1.   |---key=A------------->|                       |
 2.   |                      |<-DTLS-SRTP handshake->|
 3.   |<--key=B--------------|                       |
 4.   |                      |--new_srtp_key:A------>|
 5.   |                      |<--SRTP, encrypted w/B-|
 5.   |<-SRTP, encrypted w/B-|                       |
 6.   |-SRTP, encrypted w/A->|                       |
 7.   |                      |-SRTP, encrypted w/A-->|
      |                      |                       |]]></artwork>

        <postamble></postamble>
      </figure>

      <t>The message flow is as follows (similar steps occur with SRTCP):<list
          style="numbers">
          <t>Security Descriptions endpoint discloses its SRTP key to the SBC
          (a=crypto).</t>

          <t>SBC completes DTLS-SRTP handshake. From this handshake, the SBC
          derives the SRTP key for traffic from the DTLS-SRTP endpoint (key B)
          and to the DTLS-SRTP endpoint (key C).</t>

          <t>The SBC communicates the SRTP encryption key (key B) to the
          Security Descriptions endpoint.</t>

          <t>The SBC uses the EKT to indicate that SRTP packets will be
          encrypted with 'key A' towards the DTLS-SRTP endpoint.</t>

          <t>The DTLS-SRTP endpoint sends an SRTP key, encrypted with its key
          B. This is received by the SBC.</t>

          <t>The received SRTP packet is simply forwarded; the SBC does not
          need to do anything with this packet as its key (key B) was
          communicated in step 3.</t>

          <t>The Security Descriptions endpoint sends an SRTP packet,
          encrypted with its key A.</t>

          <t>The received SRTP packet is simply forwarded; the SBC does not
          need to do anything with this packet as its key (key A) was
          communicated in step 4.</t>
        </list></t>
    </section>
  </back>
</rfc>

PAFTECH AB 2003-20262026-04-23 20:47:08