[Docs] [txt|pdf] [Tracker] [WG] [Email] [Diff1] [Diff2] [Nits]

Versions: 00 01

Network Working Group                                       W. Eddy, Ed.
Internet-Draft                                                   Verizon
Intended status: Informational                               S. Bellovin
Expires: January 10, 2008                            Columbia University
                                                                J. Touch
                                                               R. Bonica
                                                        Juniper Networks
                                                            July 9, 2007

   Problem Statement and Requirements for a TCP Authentication Option

Status of this Memo

   By submitting this Internet-Draft, each author represents that any
   applicable patent or other IPR claims of which he or she is aware
   have been or will be disclosed, and any of which he or she becomes
   aware will be disclosed, in accordance with Section 6 of BCP 79.

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

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

   The list of current Internet-Drafts can be accessed at

   The list of Internet-Draft Shadow Directories can be accessed at

   This Internet-Draft will expire on January 10, 2008.

Copyright Notice

   Copyright (C) The IETF Trust (2007).

Eddy, et al.            Expires January 10, 2008                [Page 1]

Internet-Draft             TCP Authentication                  July 2007


   The TCP-MD5 option is commonly used to secure BGP sessions between
   routers, although it is known to have many serious deficiencies.
   This memo presents requirements for a TCP segment authentication
   mechanism that is intended to replace TCP-MD5.  While TCP-MD5 was
   designed to protect TCP sessions whose payload is BGP, the
   applicability of the mechanism described herein is broader.  This
   mechanism can be applied to any TCP connection, regardless of

Table of Contents

   1.  Contributors . . . . . . . . . . . . . . . . . . . . . . . . .  3
   2.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  4
   3.  Problem Statement  . . . . . . . . . . . . . . . . . . . . . .  5
   4.  Requirements . . . . . . . . . . . . . . . . . . . . . . . . .  8
     4.1.  Distinguishing Requirements  . . . . . . . . . . . . . . .  8
     4.2.  Expected Constraints . . . . . . . . . . . . . . . . . . . 13
   5.  Security Considerations  . . . . . . . . . . . . . . . . . . . 15
   6.  IANA Considerations  . . . . . . . . . . . . . . . . . . . . . 16
   7.  Informative References . . . . . . . . . . . . . . . . . . . . 17
   Appendix A.  Un-Agreed Properties  . . . . . . . . . . . . . . . . 20
   Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 21
   Intellectual Property and Copyright Statements . . . . . . . . . . 22

Eddy, et al.            Expires January 10, 2008                [Page 2]

Internet-Draft             TCP Authentication                  July 2007

1.  Contributors

   This document resulted from the discussions of several IETF
   participants, including significant input from a design team within
   the TCPM working group who included (alphabetically):

      Mark Allman (mallman@icir.org)

      Steve Bellovin (smb@cs.columbia.edu)

      Ron Bonica (rbonica@juniper.net)

      Wesley Eddy (weddy@grc.nasa.gov)

      Andrew Lange (andrew.lange@alcatel.com)

      Allison Mankin (mankin@psg.com)

      Sandy Murphy (sandy@tislabs.com)

      Joe Touch (touch@isi.edu)

      Sriram Viswanathan (sriram_v@cisco.com)

      Brian Weis (bew@cisco.com)

Eddy, et al.            Expires January 10, 2008                [Page 3]

Internet-Draft             TCP Authentication                  July 2007

2.  Introduction

   Putting a security service into the transport layer has a long
   history.  SP4 [SP4] [SP4P] provided that service for the Secure Data
   Network System (SDNS); OSI incorporated SP4 into its protocol suite
   as the Transport Layer Security Protocol (TLSP) [TLSP].

   TCP/IP has not had a full-fledged equivalent, though the TCP-MD5
   option [RFC2385] has served for some purposes.  TCP-MD5 is now known
   to have several problems.  In this memo, we analyze the need for a
   TCP-based security service in Section 3 and discuss the requirements
   that a solution should meet in Section 4,

   Note that we have deliberately used the phrase "security service".
   The in-use TCP-MD5 provides authentication-only and no TCP-based
   confidentiality mechanism is deployed or yet defined by the IETF.
   This document focuses on the requirements for an authentication-only
   service to replace TCP-MD5.  If a TCP-based confidentiality service
   is also warranted, it could share many of these requirements, but
   this is beyond the scope of the current work or expressed needs from
   the community.

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   document are to be interpreted as described in [RFC2119].

Eddy, et al.            Expires January 10, 2008                [Page 4]

Internet-Draft             TCP Authentication                  July 2007

3.  Problem Statement

   The TCP-MD5 mechanism described in [RFC2385], includes a Message
   Authentication Code (MAC) in each TCP header.  The MAC value is
   computed by hashing over:

   o  the TCP pseudo-header

   o  the TCP header, excluding options, and assuming a checksum of zero

   o  the TCP segment data

   o  an independently-specified shared key or password

   To successfully spoof segments to a connection using the scheme
   described above, an attacker would not only have to guess TCP
   sequence numbers, but would also have to obtain the key that was used
   to calculate the MAC.  This key never appears in the connection

   [RFC3562] addresses key management considerations for TCP-MD5.  Based
   upon the cryptographic strength of the MD5 hashing algorithm, RFC
   3562 recommends that keys be changed at least every 90 days.
   Unfortunately, TCP-MD5 only permits keys to be changed during the
   lifetime of a TCP connection if the change is synchronized at both
   ends.  This limitation has proven to be a deterrent to the effective
   deployment of TCP-MD5, and necessitates a heuristic for key change

   Also, TCP-MD5 is entirely dependent on the MD5 hash algorithm, for
   which there are now well-known collision-finding methods.  In
   addition, the particular keyed-hash MAC construction used by TCP-MD5
   has serious cryptographic weaknesses.  An attacker who can find a
   collision in the underlying hash function can forge a MAC using a
   simple chosen-message attack.  It is quite clear that the existing
   TCP-MD5 mechanism is inadequate [I-D.manral-rpsec-existing-crypto].
   It is cryptographically unsound, requiring a process waiver to permit
   its continued use with BGP [RFC4278].

   TCP-MD5 has also been accused of not meeting operator requirements,
   even though it was originally intended for operators to protect TCP-
   based routing protocol sessions with (e.g.  BGP, and now also LDP).
   TCP-MD5 is said to have high CPU utilization.  The impact of MD5
   itself is known [RFC1810], but for specific TCP-MD5 implementations,
   hard data on the protocol's performance has not been made availble,
   nor have direct comparisons between TCP-MD5 and IPsec AH performance.

   The key management and key change synchronization difficulties

Eddy, et al.            Expires January 10, 2008                [Page 5]

Internet-Draft             TCP Authentication                  July 2007

   mentioned above have also been raised as operator concerns.  It has
   been admitted that many operators simply do not change keys on a
   regular or systematic basis, but it is not clear whether this is a
   symptom of TCP-MD5's lack of capabilities, or unrelated operational
   culture.  Based on the importance to the Internet of security for
   routing protocol sessions, it is clear that TCP-MD5 should be
   improved upon, and it seems likely that an improved version could
   greatly increase the use of TCP-based authentication for routing
   protocols and thus the robustness of routing sessions against the
   known attacks targeting TCP connections [I-D.ietf-tcpm-tcp-antispoof]

   It is less clear why authentication is needed at all within TCP
   implementations.  IPsec [RFC4301] can protect the entire TCP header
   and payload, and TLS [RFC4346] can protect the payload data within a
   TCP connection, when used by an application.  However, these existing
   solutions have their own deficiencies [I-D.ietf-tcpm-tcp-antispoof].

   The most serious problem with IPsec is that it is hard to protect an
   individual TCP connection with it [I-D.bellovin-useipsec], due to the
   lack of an API that an application can request IPsec protection for a
   specific connection via.  IPsec operates at the IP layer (with only a
   sprinkling of transport layer concepts, such as port numbers used
   within traffic selectors), and has no notion of individual transport
   layer connections and their duration (only quintuples of IP
   addresses, protocol, and port numbers), so "latching" a particular
   TCP connection to an IPsec Security Association with a corresponding
   lifetime is difficult [I-D.ietf-btns-connection-latching].

   IPsec also has problems with NAT traversal [RFC2709] [RFC3715]
   [RFC3947] [RFC3948]: NAT boxes can neither examine nor modify port
   numbers on most IPsec-protected traffic, which causes very real
   problems in many environments (though not, admittedly, when
   protecting BGP).  The net result is that IPsec usage is largely
   limited to virtual private network scenarios; it is rarely used or
   usable for individual applications over the Internet.  At this point,
   the primary use of the existing TCP-based security method is
   protecting BGP sessions between routers.  BGP speakers will rarely,
   if ever, be behind NATs, so it would seem that IPsec could be
   feasible in this use case.  The existing TCP-MD5 is similarly
   hindered by the presence of NATs.  The improved TCP authentication
   mechanism is intended for general use, not limited to BGP
   connections.  On a per-connection configurable basis, compatibility
   with NATs is a goal of this work.

   IPsec tunnels and ESP in transport or tunnel mode have often been
   criticized for their interference with firewalls and with traffic
   engineering, because they can hide port numbers and flags.  A TCP

Eddy, et al.            Expires January 10, 2008                [Page 6]

Internet-Draft             TCP Authentication                  July 2007

   security option might choose to expose such fields for examination.

   TLS does not suffer from any of these afflictions; however, it poses
   issues of its own.  The integrated key management in TLS works well
   in many environments, but is too heavy-weight or otherwise
   inappropriate for others.  A more serious issue is the limited scope
   of protection provided by TLS.  It operates strictly above TCP, and
   thus provides no protection at all against attacks against the TCP
   header itself.  Even if TLS is in use, it is possible for attackers
   to reset connections (US-CERT Advisory TA04-111A) or perpetrate other
   mischief that affects the TCP connection state before TLS processing
   occurs [I-D.ietf-tcpm-tcp-antispoof].

   Since TCP-MD5 is deeply flawed and neither IPsec nor TLS currently
   provide the desired granularity of protection for some uses, it is
   clear that an intermediate protection mechanism can be justfied.
   There have been multiple proposals presented recently to fill this
   void [I-D.bonica-tcp-auth] [I-D.weis-tcp-auth-auto-ks]
   [I-D.touch-tcpm-tcp-simple-auth], but without an agreed-upon set of
   requirements, evaluating these proposals has been postponed.  In
   Section 4 within this document, we provide a set of requirements that
   has been agreed upon by authors of all of the currently known
   proposed solutions.

Eddy, et al.            Expires January 10, 2008                [Page 7]

Internet-Draft             TCP Authentication                  July 2007

4.  Requirements

   In this section, we present the distinguishing requirements for a
   future TCP security option, based on a consensus within the TCPM
   Authentication Option design team.  These requirements are intended
   to be used as a means of evaluating potential solutions.  These
   requirements partially have some basis in [RFC4808], and also have
   some commonality with other requirement sets developed for BGP
   session security [I-D.behringer-bgp-session-sec-req].  We also
   include some expected constraints or behaviors of a solution in
   Section 4.2, that are not expected to be useful in evaluating between
   differing approaches, but are refinements that could be compatible
   with any solution approach.  Some suggested properties that the
   design team was not able to obtain a consensus for or against are
   listed in Appendix A.

4.1.  Distinguishing Requirements

   The requirements that a solution must fulfill are:

   1.  Protected Elements:

       A.  TCP Pseudoheader

           The pseudoheader of specific IPv4 or IPv6 fields used in the
           computation of a segment's TCP checksum, from [RFC0793] and
           [RFC2460], is protected.  By including source and destination
           IP addresses, this influences operation through NATs in a
           similar way to IPsec's Authentication Header, so although
           pseudoheader coverage MUST be possible in any viable
           solution, it MUST also be optional on a per-connection basis,

           For checksum purposes, the header of a TCP connection is the
           combination of its TCP Pseudoheader and its TCP Header.  The
           IP addresses of the pseudoheader are included because they
           (together with the port numbers) define the connection; other
           fields are included to protect fields of the IP header that
           otherwise affect the TCP connection (in the latter case,
           largely by their inclusion in the TCP checksum).

       B.  Base TCP Header

           The full base TCP header is protected, excluding any TCP
           options and the TCP checksum.  By covering TCP port numbers,
           this influences operation through NATs in a similar way to
           IPsec's Authentication Header, so while port number coverage
           MUST be possible in any viable solution, it also MUST be
           optional on a per-connection basis.

Eddy, et al.            Expires January 10, 2008                [Page 8]

Internet-Draft             TCP Authentication                  July 2007

           The TCP Header is included by definition, since the purpose
           of this security option is to protect the TCP header.

       C.  TCP Options

           Additionally, each defined TCP option type may be either
           selected for or excluded from protection.  This is configured
           on a per-option type, per-connection basis, and is static for
           the lifetime of a TCP connection.

           Other TCP options may or may not be protected by this
           security option, as desired.  The primary reason for
           excluding options is efficiency, and because this level of
           protection can be relaxed in a way that impacts only an
           individual connection, this is a user choice.

           Note that the authentication option itself MUST be included,
           with the authentication hash zeroed out.

       D.  TCP Data

           The payload of each TCP segment containing the data given to
           applications MUST also be protected.

   2.  Option Structure Requirements:

       A.  Privacy

           The authentication option MUST NOT directly expose sensitive
           security parameters, so that a third party's ability to view
           packets does not also permit them to inject authenticable
           packets or to otherwise determine information that could be
           used to compromise a particular connection, or other
           connections, between a pair of hosts.

       B.  Allow Optional

           A host capable of parsing the authentication option MUST be
           able to requrie or ignore the option on received segments on
           a per-connection basis.

           The purposed of the option is to authenticate connections;
           hosts must be able to discard segments sent to connections
           intended to be authenticated (i.e. they MUST be able to
           require the option's use).  Authentication determines the ID
           of the source of a packet; some hosts may not be interested
           in verifying the ID.  Presumably, use of the option would be
           determined a-priori, before a connection is established by a

Eddy, et al.            Expires January 10, 2008                [Page 9]

Internet-Draft             TCP Authentication                  July 2007

           separate key and/or policy management system, but it still
           may be useful to offload or otherwise ignore an expensive
           authentication calculation, especially if the resulting ID
           confirmation is not desired.

       C.  Require Non-Optional

           A host capable of sending the authentication option MUST be
           able to coordinate in-band whether the option should be
           required or might be ignored for a particular connection with
           a capable receiver.

           This requirement supports senders who prefer to use the
           option, but who are also willing to support hosts not
           implementing the option.  Such coordination would typically
           happen in the key management system, but since that system
           could be manual, an in-band mechanism to confirm use of this
           option and backoff if not supported is required.  This
           mechanism would also prevent backoff if the sender does not
           desire that behavior.

       D.  Standard Parsing

           The authentication option MUST be trivially parseable by
           those TCP implementations that do not support it.  This means
           that it must follow the [RFC0793] format of including a type
           and length field, so that it can be skipped over when it is
           not supported by an implementation.  TCP already specifies
           that hosts not supporting an option ignore that option in
           received segments; stating this requirement here simply
           ensures that TCP authentication solutions do not alter the
           format of the base TCP header or radically depart from the
           typical options encoding.

       E.  Compatible with Large Windows

           The authentication option MUST allow the concurrent use of
           timestamps and window-scaling within protected connections,
           as excluding these could limit its range of performance.

           These options are in common use, and are needed for
           performance over high-speed or high-delay paths.  Use of the
           authentication option thus needs to permit the use of these
           options, or its practical deployability will be severely

       F.  Compatible with SACK

Eddy, et al.            Expires January 10, 2008               [Page 10]

Internet-Draft             TCP Authentication                  July 2007

           If the use of Selective Acknowledgements (SACK) is negotiated
           on a connection, the authentication option MUST allow room
           for at least one SACK block to be included in the TCP
           options, and preferably more.

           This option, like (E), is in common use, and is needed for
           performance in large-window, lossy connections.  Use of the
           authentication option thus needs to permit the use of SACK.

   3.  Cryptography Requirements

       A.  Baseline Defaults

           There MUST be at least one set of particular cryptography
           algorithms or constructions whose use is supported by all
           implementations and can be safely assumed to be supported by
           any implementation of the authentication option.

           This requirement is intended to support interoperability of
           this option, by having a single default.

       B.  Good Algorithms and Constructions

           The authentication option MUST support default cryptography
           algorithms and constructions that are accepted by the
           community.  This means it MUST NOT rely on non-standard or
           ad-hoc hash functions, keyed-hash constructions, signature
           schemes, or other functions, and MUST use published and
           standard schemes (i.e. it should use a construction like HMAC
           versus the form of keyed-hash used in TCP-MD5).

           This requirement is intended to correct the flaws in the
           strength of authentication provided by the keyed hash used in

       C.  Algorithm Agility

           The authentication option MUST be capable of supporting
           algorithms other than its defaults, in order to adapt to
           future discoveries.  An implementation that supports multiple
           algorithms MUST permit concurrent connections to use
           different algorithms.

           The existing TCP-MD5 requires substantial revision or
           retirement because its algorithms cannot be replaced.  This
           requirement allows the authentication option to be agile to
           algorithmic attacks, where additional algorithms can be added
           as needed.

Eddy, et al.            Expires January 10, 2008               [Page 11]

Internet-Draft             TCP Authentication                  July 2007

       D.  Order-Independent Processing

           Authentication MUST be performed on individual, unordered TCP
           segments, so that it is not severely influenced by reasonable
           amounts of packet loss or reordering.

           TCP headers are processed in the order received, although the
           data is reordered based on header information.  As a result,
           header fields must be authenticated in the order received; to
           reorder them first would alter TCP semantics, and would
           potentially require data in unauthenticated segments to be
           quarantined (i.e. copied again) until authenticated later.

       E.  Parameter Changes Require Key Changes

           A change in the keys used MUST accompany any change in the
           other parameters the cryptography functions for the
           authentication option are configured with.

           This requirement allows the design of a compact option.  It
           allows the key ID and key itself to indicate the parameters,
           rather than requiring header fields for them.  It also avoids
           interpreting those parameters from in-band information,
           further avoiding exposing them to parties on the path.

   4.  Keying Requirements

       A.  Intraconnection Rekeying

           Within the course of a single connection, the authentication
           option MUST accomodate rekeying.

           TCP spoofing attacks, which this option is intended to
           defeat, are often targeted at relatively long-lived
           connections.  Use of a single key over a long connection is a
           known security problem, so it would be preferable to either
           limit the length of a connection or require in-band keying

           Unfortunately, not all applications are easy to restart.
           BGP, for which this option is intended, is being augmented
           for graceful restart [RFC4724] [RFC4781], but this extension
           is under recent scrutiny.  TCP itself has no limit on the
           length of a connection, and it would be preferable to avoid
           modifying this semantic.

       B.  Efficient Rekeying

Eddy, et al.            Expires January 10, 2008               [Page 12]

Internet-Draft             TCP Authentication                  July 2007

           A rekeying event MUST NOT significantly affect performance of
           the TCP connection.  Most segments should be validated by a
           single pass of the construction of cryptography algorithms
           used for authentication, and no validations should require
           more than a small, fixed number of passes.

           Any aspect of this option which is inefficient is likely to
           inhibit its deployment.  When using this option, segments may
           arrive out of order, and it would be inefficient to determine
           which key is appropriate via a large number of trials.  Such
           trials would present a DoS vulnerability during rekeying.
           This issue is discussed in [RFC4808].

       C.  Automated and Manual Keying

           The authentication option MUST support both manual
           configuration of preshared keys and automated key management.
           This allows for different modes of operation depending on the
           user's particular deployment environment.

       D.  Key-Mananagement Agnostic

           The per-segment authentication is performed without regards
           to the manner in which keying material is obtained.  This
           requirement decouples the option mechanism itself from the
           key management system used, so that either multiple protocols
           can be integrated, or any flawed methods can be easily
           replaced in the future.

4.2.  Expected Constraints

   In addition to having a wire format that supports the Distinguishing
   Requirements, a solution should include the following caveats in its
   internal operation.

   1.  Silent Failure

       On failure (due to incorrect or missing authentication data),
       segments MUST be silently discarded, with no reply generated.
       Shuch events SHOULD be logged periodically.  Failed segments MUST
       NOT alter the protocol state of the TCP connection itself.

       Silence and the use of only periodic logging prevents the
       creation of a new DoS opportunity.

   2.  Maximum of One Option per Segment

       At most, one authentication option MUST be allowed per segment.

Eddy, et al.            Expires January 10, 2008               [Page 13]

Internet-Draft             TCP Authentication                  July 2007

       The presence of multiple options MUST be treated as a failure.

       Use of multiple options would present another DoS opportunity,
       and provides no additional protection vs. a single option with
       appropriate connection latching to other mechanisms, if desired.

   3.  Outgoing All-or-None Operation

       Within a connection, once the authentication option is enabled,
       all segments MUST carry the option.

       This prevents headers and/or data from being injected into a
       protected connection.

   4.  Incoming All-Checked Operation

       An implementation capable of using the authentication option MUST
       check every incoming segment's connection state to decide whether
       the option's presence is required.

       This requirement allows a host to determine which connections
       require the option, vs. which allow it as optional.  Checking
       connection state for every incoming segment enforces required use
       for indicated connections.

   5.  Non-Interaction with TCP-MD5

       An implementation MUST NOT allow a connection to simultaneously
       use the new authentication option and TCP-MD5.  An implementation
       MAY support the use of either exclusively the new authentication
       option or exclusively TCP-MD5 for each individual connection.

       This option is intended to supercede TCP-MD5, and in the spirit
       of (2) above, only one such option is useful per connection.
       Support for existing TCP-MD5 would support legacy interoperation.

   6.  Optional ICMP Discard

       An implementation MUST be configurable to allow a protected
       connection to ignore incoming ICMP Type 3 messages with Codes
       2-4.  This SHOULD be the default configuration.

       This requirement prevents an ICMP attack on protected connections
       via unprotected/unauthenticable (ICMP) packets.

Eddy, et al.            Expires January 10, 2008               [Page 14]

Internet-Draft             TCP Authentication                  July 2007

5.  Security Considerations

   This document does not specify any protocol; it discusses known
   security problems with a currently deployed protocol, and the
   requirements for fixing those problems in a new protocol.  This
   document is itself a set of security considerations, and its
   publication raises no new security considerations.

Eddy, et al.            Expires January 10, 2008               [Page 15]

Internet-Draft             TCP Authentication                  July 2007

6.  IANA Considerations

   This document does not update or create any IANA registries.

Eddy, et al.            Expires January 10, 2008               [Page 16]

Internet-Draft             TCP Authentication                  July 2007

7.  Informative References

              Behringer, M., "BGP Session Security Requirements",
              draft-behringer-bgp-session-sec-req-01 (work in progress),
              May 2007.

              Bellovin, S., "Guidelines for Mandating the Use of IPsec",
              draft-bellovin-useipsec-06 (work in progress),
              February 2007.

              Bonica, R., "Authentication for TCP-based Routing and
              Management Protocols", draft-bonica-tcp-auth-06 (work in
              progress), February 2007.

              Williams, N., "IPsec Channels: Connection Latching",
              draft-ietf-btns-connection-latching-01 (work in progress),
              March 2007.

              Touch, J., "Defending TCP Against Spoofing Attacks",
              draft-ietf-tcpm-tcp-antispoof-06 (work in progress),
              February 2007.

              Ramaiah, A., "Improving TCP's Robustness to Blind In-
              Window Attacks", draft-ietf-tcpm-tcpsecure-07 (work in
              progress), February 2007.

              Manral, V., "Issues with existing Cryptographic Protection
              Methods for Routing  Protocols",
              draft-manral-rpsec-existing-crypto-04 (work in progress),
              April 2007.

              Touch, J. and A. Mankin, "The TCP Simple Authentication
              Option", Internet-Draft
              draft-touch-tcpm-tcp-simple-auth-02 (work in progress),,
              October 2006.

              Weis, B., "Automated key selection extension for the TCP
              Enhanced Authentication  Option",
              draft-weis-tcp-auth-auto-ks-02 (work in progress),

Eddy, et al.            Expires January 10, 2008               [Page 17]

Internet-Draft             TCP Authentication                  July 2007

              March 2007.

   [RFC0793]  Postel, J., "Transmission Control Protocol", STD 7,
              RFC 793, September 1981.

   [RFC1810]  Touch, J., "Report on MD5 Performance", RFC 1810,
              June 1995.

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

   [RFC2385]  Heffernan, A., "Protection of BGP Sessions via the TCP MD5
              Signature Option", RFC 2385, August 1998.

   [RFC2460]  Deering, S. and R. Hinden, "Internet Protocol, Version 6
              (IPv6) Specification", RFC 2460, December 1998.

   [RFC2709]  Srisuresh, P., "Security Model with Tunnel-mode IPsec for
              NAT Domains", RFC 2709, October 1999.

   [RFC3562]  Leech, M., "Key Management Considerations for the TCP MD5
              Signature Option", RFC 3562, July 2003.

   [RFC3715]  Aboba, B. and W. Dixon, "IPsec-Network Address Translation
              (NAT) Compatibility Requirements", RFC 3715, March 2004.

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

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

   [RFC4107]  Bellovin, S. and R. Housley, "Guidelines for Cryptographic
              Key Management", BCP 107, RFC 4107, June 2005.

   [RFC4278]  Bellovin, S. and A. Zinin, "Standards Maturity Variance
              Regarding the TCP MD5 Signature Option (RFC 2385) and the
              BGP-4 Specification", RFC 4278, January 2006.

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

   [RFC4346]  Dierks, T. and E. Rescorla, "The Transport Layer Security
              (TLS) Protocol Version 1.1", RFC 4346, April 2006.

   [RFC4724]  Sangli, S., Chen, E., Fernando, R., Scudder, J., and Y.

Eddy, et al.            Expires January 10, 2008               [Page 18]

Internet-Draft             TCP Authentication                  July 2007

              Rekhter, "Graceful Restart Mechanism for BGP", RFC 4724,
              January 2007.

   [RFC4781]  Rekhter, Y. and R. Aggarwal, "Graceful Restart Mechanism
              for BGP with MPLS", RFC 4781, January 2007.

   [RFC4808]  Bellovin, S., "Key Change Strategies for TCP-MD5",
              RFC 4808, March 2007.

   [SP4]      Dinkel, C., "Secure Data Network System (SDNS) Network,
              Transport, and Message Security Protocols", NISTIR 90-
              4250, 1990.

   [SP4P]     Branstad, D., Dorman, J., Housley, R., and J. Randall,
              "SP4: A Transport Encapsulation Security Protocol", Third
              Aerospace Security Conference Proceedings, December 1987.

   [TLSP]     "Information Technology -- Telecommunications and
              Information Exchange Between systems -- Transport Layer
              Security Protocol", ISO/IEC 10736, 1995.

Eddy, et al.            Expires January 10, 2008               [Page 19]

Internet-Draft             TCP Authentication                  July 2007

Appendix A.  Un-Agreed Properties

   There were some items that were suggested as requirements but which
   were not ratified by all participants in the design team.  These are
   listed here.

   1.  Saves Work When Optional

       A host sending TCP segments should be able to detect on a per-
       connection basis whether the authentication option is required or
       is being ignored by a receiver who supports the option.

   2.  Single-Pass Rekeying

       The authentication option should support rekeying where incoming
       segments are validated using a single pass of the cryptographic
       construction used for authentication.

Eddy, et al.            Expires January 10, 2008               [Page 20]

Internet-Draft             TCP Authentication                  July 2007

Authors' Addresses

   Wesley M. Eddy (editor)
   Verizon Federal Network Systems
   NASA Glenn Research Center
   21000 Brookpark Rd, MS 54-5
   Cleveland, OH  44135

   Phone: 216-433-6682
   Email: weddy@grc.nasa.gov

   Steven M. Bellovin
   Columbia University
   1214 Amsterdam Avenue
   MC 0401
   New York, NY  10027

   Phone: +1 212 939 7149
   Email: bellovin@acm.org

   Joe Touch
   4676 Admirality Way
   Marina del Rey, CA  90292-6695

   Phone: +1 (310) 448-9151
   Email: touch@isi.edu

   Ronald P. Bonica
   Juniper Networks
   2251 Corporate Park Drive
   Herndon, VA  20171

   Email: rbonica@juniper.net

Eddy, et al.            Expires January 10, 2008               [Page 21]

Internet-Draft             TCP Authentication                  July 2007

Full Copyright Statement

   Copyright (C) The IETF Trust (2007).

   This document is subject to the rights, licenses and restrictions
   contained in BCP 78, and except as set forth therein, the authors
   retain all their rights.

   This document and the information contained herein are provided on an

Intellectual Property

   The IETF takes no position regarding the validity or scope of any
   Intellectual Property Rights or other rights that might be claimed to
   pertain to the implementation or use of the technology described in
   this document or the extent to which any license under such rights
   might or might not be available; nor does it represent that it has
   made any independent effort to identify any such rights.  Information
   on the procedures with respect to rights in RFC documents can be
   found in BCP 78 and BCP 79.

   Copies of IPR disclosures made to the IETF Secretariat and any
   assurances of licenses to be made available, or the result of an
   attempt made to obtain a general license or permission for the use of
   such proprietary rights by implementers or users of this
   specification can be obtained from the IETF on-line IPR repository at

   The IETF invites any interested party to bring to its attention any
   copyrights, patents or patent applications, or other proprietary
   rights that may cover technology that may be required to implement
   this standard.  Please address the information to the IETF at


   Funding for the RFC Editor function is provided by the IETF
   Administrative Support Activity (IASA).

Eddy, et al.            Expires January 10, 2008               [Page 22]

Html markup produced by rfcmarkup 1.107, available from http://tools.ietf.org/tools/rfcmarkup/