Network Working Group                                        S. Bellovin
Request for Comments: 4808                           Columbia University
Category: Informational                                       March 2007


                  Key Change Strategies for TCP-MD5

Status of This Memo

  This memo provides information for the Internet community.  It does
  not specify an Internet standard of any kind.  Distribution of this
  memo is unlimited.

Copyright Notice

  Copyright (C) The IETF Trust (2007).

Abstract

  The TCP-MD5 option is most commonly used to secure BGP sessions
  between routers.  However, changing the long-term key is difficult,
  since the change needs to be synchronized between different
  organizations.  We describe single-ended strategies that will permit
  (mostly) unsynchronized key changes.



























Bellovin                     Informational                      [Page 1]

RFC 4808                   TCP-MD5 Key Change                 March 2007


1.  Introduction

  The TCP-MD5 option [RFC2385] is most commonly used to secure BGP
  sessions between routers.  However, changing the long-term key is
  difficult, since the change needs to be synchronized between
  different organizations.  Worse yet, if the keys are out of sync, it
  may break the connection between the two routers, rendering repair
  attempts difficult.

  The proper solution involves some sort of key management protocol.
  Apart from the complexity of such things, RFC 2385 was not written
  with key changes in mind.  In particular, there is no KeyID field in
  the option, which means that even a key management protocol would run
  into the same problem.

  Fortunately, a heuristic permits key change despite this protocol
  deficiency.  The change can be installed unilaterally at one end of a
  connection; it is fully compatible with the existing protocol.

1.1.  Terminology

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

2.  The Algorithm

  Separate algorithms are necessary for transmission and reception.
  Reception is easier; we explain it first.

2.1.  Reception

  A receiver has a list of valid keys.  Each key has a (conceptual)
  timestamp associated with it.  When a segment arrives, each key is
  tried in turn.  The segment is discarded if and only if it cannot be
  validated by any key in the list.

  In principle, there is no need to test keys in any particular order.
  For performance reasons, though, a simple most-recently-used (MRU)
  strategy -- try the last valid key first -- should work well.  More
  complex mechanisms, such as examining the TCP sequence number of an
  arriving segment to see whether it fits in a hole, are almost
  certainly unnecessary.  On the other hand, validating that a received
  segment is putatively legal, by checking its sequence number against
  the advertised window, can help avoid denial of service attacks.

  The newest key that has successfully validated a segment is marked as
  the "preferred" key; see below.



Bellovin                     Informational                      [Page 2]

RFC 4808                   TCP-MD5 Key Change                 March 2007


  Implicit in this scheme is the assumption that older keys will
  eventually be unneeded and can be removed.  Accordingly,
  implementations SHOULD provide an indication of when a key was last
  used successfully.

2.2.  Transmission

  Transmission is more complex, because the sender does not know which
  keys can be accepted at the far end.  Accordingly, the conservative
  strategy is to delay using any new keys for a considerable amount of
  time, probably measured in days.  This time interval is the amount of
  asynchronicity the parties wish to permit; it is agreed upon out of
  band and configured manually.

  Some automation is possible, however.  If a key has been used
  successfully to validate an incoming segment, clearly the other side
  knows it.  Accordingly, any key marked as "preferred" by the
  receiving part of a stack SHOULD be used for transmissions.

  A sophisticated implementation could try alternate keys if the TCP
  retransmission counter gets too high.  (This is analogous to dead
  gateway detection.)  In particular, if a key change has just been
  attempted but such segments are not acknowledged, it is reasonable to
  fall back to the previous key and issue an alert of some sort.
  Similarly, an implementation with a new but unused key could
  occasionally try to use it, much in the way that TCP implementations
  probe closed windows.  Doing this avoids the "silent host" problem
  discussed in Section 3.1.  This should be done at a moderately slow
  rate.

  Note that there is an ambiguity when an acknowledgment is received
  for a segment transmitted with two different keys.  The TCP Timestamp
  option [RFC1323] can be used for disambiguation.

3.  Operations

3.1.  Single-Ended Operations

  Suppose only one end of the connection has this algorithm
  implemented.  The new key is provisioned on that system, with a start
  time far in the future -- sufficiently far, in fact, that it will not
  be used spontaneously.  After the key is ready, the other end is
  notified, out-of-band, that a key change can commence.

  At some point, the other end is upgraded.  Because it does not have
  multiple keys available, it will start using the new key immediately
  for its transmission, and will drop all segments that use the old
  key.  As soon as it tries to transmit, the upgraded side will



Bellovin                     Informational                      [Page 3]

RFC 4808                   TCP-MD5 Key Change                 March 2007


  designate the new key as preferred, and will use it for all of its
  transmissions.  Note specifically that this will include
  retransmissions of any segments rejected because they used the old
  key.

  There is a problem if the unchanged machine is a "silent host" -- a
  host that has nothing to say, and hence does not transmit.  The best
  way to avoid this is for an upgraded machine to try a variety of keys
  in the event of repeated unacknowledged packets, and to probe for new
  unused keys during silent periods, as discussed in Section 2.2.
  Alternatively, application-level KeepAlive messages may be used to
  ensure that neither end of the connection is completely silent.  See,
  for example, Section 4.4 of [RFC4271] or Section 3.5.4 of [RFC3036].

3.2.  Double-Ended Operations

  Double-ended operations are similar, save that both sides deploy the
  new key at about the same time.  One should be configured to start
  using the new key at a point where it is reasonably certain that the
  other side would have it installed, too.  Assuming that has in fact
  happened, the new key will be marked "preferred" on both sides.

3.3.  Monitoring

  As noted, implementations should monitor when a key was last used for
  transmission or reception.  Any monitoring mechanism can be used;
  most likely, it will be one or both of a MIB object or objects and
  the vendor's usual command-line mechanism for displaying data of this
  type.  Regardless, the network operations center should keep track of
  this.  When a new key has been used successfully for both
  transmission and reception for a reasonable amount of time -- the
  exact value isn't crucial, but it should probably be longer than
  twice the maximum segment lifetime -- the old key can be marked for
  deletion.  There is an implicit assumption here that there will not
  be substantial overlap in the usage period of such keys; monitoring
  systems should look for any such anomalies, of course.

4.  Moving Forward

  As implied in Section 1, this is an interim strategy, intended to
  make TCP-MD5 operationally usable today.  We do not suggest or
  recommend it as a long-term solution.  In this section, we make some
  suggestions about the design of a future TCP authentication option.

  The first and most obvious change is to replace keyed MD5 with a
  stronger MAC [RFC4278].  Today, HMAC-SHA1 [RFC4634] is the preferred
  choice, though others such as UMAC [RFC4418] should be considered as
  well.



Bellovin                     Informational                      [Page 4]

RFC 4808                   TCP-MD5 Key Change                 March 2007


  A new authentication option should contain some form of a Key ID
  field.  Such an option would permit unambiguous identification of
  which key was used to create the MAC for a given segment, sparing the
  receiver the need to engage in the sort of heuristics described here.
  A Key ID is useful with both manual and automatic key management.
  (Note carefully that we do not prescribe any particular Key ID
  mechanism here.  Rather, we are stating a requirement: there must be
  a simple, low-cost way to select a particular key, and it must be
  possible to rekey without tearing down long-lived connections.)

  Finally, an automated key management mechanism should be defined.
  The general reasoning for that is set forth in [RFC4107]; specific
  issues pertaining to BGP and TCP are given in [RFC3562].

5.  Security Considerations

  In theory, accepting multiple keys simultaneously makes life easier
  for an attacker.  In practice, if the recommendations in [RFC3562]
  are followed, this should not be a problem.

  New keys must be communicated securely.  Specifically, new key
  messages must be kept confidential and must be properly
  authenticated.

  Having multiple keys makes CPU denial-of-service attacks easier.
  This suggests that keeping the overlap period reasonably short is a
  good idea.  In addition, the Generalized TTL Security Mechanism
  [RFC3682], if applicable to the local topology, can help.  Note that
  most of the time, only one key will exist; virtually all of the
  remaining time there will be only two keys in existence.

6.  IANA Considerations

  There are no IANA actions required.  The TCP-MD5 option number is
  defined in [RFC2385], and is currently listed by IANA.

7.  Acknowledgments

  I'd like to thank Ron Bonica, Randy Bush, Ross Callon, Rob Evans,
  Eric Rescorla, and Sam Weiler for their comments and inspiration.











Bellovin                     Informational                      [Page 5]

RFC 4808                   TCP-MD5 Key Change                 March 2007


8.  References

8.1.  Normative References

  [RFC1323]  Jacobson, V., Braden, B., and D. Borman, "TCP Extensions
             for High Performance", RFC 1323, May 1992.

  [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.

8.2.  Informative References

  [RFC3036]  Andersson, L., Doolan, P., Feldman, N., Fredette, A., and
             B. Thomas, "LDP Specification", RFC 3036, January 2001.

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

  [RFC3682]  Gill, V., Heasley, J., and D. Meyer, "The Generalized TTL
             Security Mechanism (GTSM)", RFC 3682, February 2004.

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

  [RFC4271]  Rekhter, Y., Li, T., and S. Hares, "A Border Gateway
             Protocol 4 (BGP-4)", RFC 4271, January 2006.

  [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.

  [RFC4418]  Krovetz, T., "UMAC: Message Authentication Code using
             Universal Hashing", RFC 4418, March 2006.

  [RFC4634]  Eastlake, D. and T. Hansen, "US Secure Hash Algorithms
             (SHA and HMAC-SHA)", RFC 4634, August 2006.












Bellovin                     Informational                      [Page 6]

RFC 4808                   TCP-MD5 Key Change                 March 2007


Author's Address

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

  Phone: +1 212 939 7149
  EMail: [email protected]








































Bellovin                     Informational                      [Page 7]

RFC 4808                   TCP-MD5 Key Change                 March 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
  "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
  OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY, THE IETF TRUST AND
  THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS
  OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF
  THE INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
  WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.

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
  http://www.ietf.org/ipr.

  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
  [email protected].

Acknowledgement

  Funding for the RFC Editor function is currently provided by the
  Internet Society.







Bellovin                     Informational                      [Page 8]