RFC 6181






Internet Engineering Task Force (IETF)                        M. Bagnulo
Request for Comments: 6181                                          UC3M
Category: Informational                                       March 2011
ISSN: 2070-1721


       Threat Analysis for TCP Extensions for Multipath Operation
                        with Multiple Addresses

Abstract



   Multipath TCP (MPTCP for short) describes the extensions proposed for
   TCP so that endpoints of a given TCP connection can use multiple
   paths to exchange data.  Such extensions enable the exchange of
   segments using different source-destination address pairs, resulting
   in the capability of using multiple paths in a significant number of
   scenarios.  Some level of multihoming and mobility support can be
   achieved through these extensions.  However, the support for multiple
   IP addresses per endpoint may have implications on the security of
   the resulting MPTCP.  This note includes a threat analysis for MPTCP.

Status of This Memo



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

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Not all documents
   approved by the IESG are a candidate for any level of Internet
   Standard; see Section 2 of RFC 5741.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   http://www.rfc-editor.org/info/rfc6181.















Bagnulo                       Informational                     [Page 1]

RFC 6181                  MPTCP Threat Analysis               March 2011


Copyright Notice



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

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (http://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents



   1.  Introduction . . . . . . . . . . . . . . . . . . . . . . . . .  3
   2.  Scope  . . . . . . . . . . . . . . . . . . . . . . . . . . . .  3
   3.  Related Work . . . . . . . . . . . . . . . . . . . . . . . . .  4
   4.  Basic MPTCP  . . . . . . . . . . . . . . . . . . . . . . . . .  6
   5.  Flooding Attacks . . . . . . . . . . . . . . . . . . . . . . .  8
   6.  Hijacking Attacks  . . . . . . . . . . . . . . . . . . . . . . 10
     6.1.  Hijacking Attacks to the Basic MPTCP . . . . . . . . . . . 10
     6.2.  Time-Shifted Hijacking Attacks . . . . . . . . . . . . . . 13
     6.3.  NAT Considerations . . . . . . . . . . . . . . . . . . . . 14
   7.  Recommendation . . . . . . . . . . . . . . . . . . . . . . . . 15
   8.  Security Considerations  . . . . . . . . . . . . . . . . . . . 16
   9.  Contributors . . . . . . . . . . . . . . . . . . . . . . . . . 16
   10. Acknowledgments  . . . . . . . . . . . . . . . . . . . . . . . 16
   11. References . . . . . . . . . . . . . . . . . . . . . . . . . . 16
     11.1. Normative References . . . . . . . . . . . . . . . . . . . 16
     11.2. Informative References . . . . . . . . . . . . . . . . . . 16


















Bagnulo                       Informational                     [Page 2]

RFC 6181                  MPTCP Threat Analysis               March 2011


1.  Introduction



   Multipath TCP (MPTCP for short) describes the extensions proposed for
   TCP [RFC0793] so that endpoints of a given TCP connection can use
   multiple paths to exchange data.  Such extensions enable the exchange
   of segments using different source-destination address pairs,
   resulting in the capability of using multiple paths in a significant
   number of scenarios.  Some level of multihoming and mobility support
   can be achieved through these extensions.  However, the support for
   multiple IP addresses per endpoint may have implications on the
   security of the resulting MPTCP.  This note includes a threat
   analysis for MPTCP.  There are many other ways to provide multiple
   paths for a TCP connection other than the usage of multiple
   addresses.  The threat analysis performed in this document is limited
   to the specific case of using multiple addresses per endpoint.

2.  Scope



   There are multiple ways to achieve Multipath TCP.  Essentially, what
   is needed is for different segments of the communication to be
   forwarded through different paths by enabling the sender to specify
   some form of path selector.  There are multiple options for such a
   path selector, including the usage of different next hops, using
   tunnels to different egress points, and so on.  The scope of the
   analysis included in this note is limited to a particular approach,
   namely MPTCP, that relies on the usage of multiple IP address per
   endpoint and that uses different source-destination address pairs as
   a means to express different paths.  So, in the rest of this note,
   the MPTCP expression will refer to this multi-addressed flavor of
   Multipath TCP [MPTCP-MULTIADDRESSED].

   This goal of this note is to perform a threat analysis for MPTCP.
   Introducing the support of multiple addresses per endpoint in a
   single TCP connection may result in additional vulnerabilities
   compared to single-path TCP.  The scope of this note is to identify
   and characterize these new vulnerabilities.  So, the scope of the
   analysis is limited to the additional vulnerabilities resulting from
   the multi-address support compared to the current TCP (where each
   endpoint only has one address available for use per connection).  A
   full analysis of the complete set of threats is explicitly out of the
   scope.  The goal of this analysis is to help the MPTCP designers
   create an MPTCP specification that is as secure as the current TCP.
   It is a non-goal of this analysis to help in the design of MPTCP that
   is more secure than regular TCP.

   The focus of the analysis is on attackers that are not along the
   path, at least not during the whole duration of the connection.  In
   the current single-path TCP, an on-path attacker can launch a



Bagnulo                       Informational                     [Page 3]

RFC 6181                  MPTCP Threat Analysis               March 2011


   significant number of attacks, including eavesdropping, connection
   hijacking Man-in-the-Middle (MiTM) attacks, and so on.  However, it
   is not possible for the off-path attackers to launch such attacks.
   There is a middle ground in case the attacker is located along the
   path for a short period of time to launch the attack and then moves
   away, but the attack effects still apply.  These are the so-called
   time-shifted attacks.  Since these are not possible in today's TCP,
   they are also consider in the analysis.  So, summarizing, both
   attacks launched by off-path attackers and time-shifted attacks are
   considered to be within scope.  Attacks launched by on-path attackers
   are out of scope, since they also apply to current single-path TCP.

   However, that some current on-path attacks may become more difficult
   with Multipath TCP, since an attacker (on a single path) will not
   have visibility of the complete data stream.

3.  Related Work



   There is a significant amount of previous work in terms of analysis
   of protocols that support address agility.  The most relevant ones
   are presented in this section.

   Most of the problems related to address agility have been deeply
   analyzed and understood in the context of Route Optimization support
   in Mobile IPv6 (MIPv6 RO) [RFC3775].  [RFC4225] includes the
   rationale for the design of the security of MIPv6 RO.  All the
   attacks described in the aforementioned analysis apply here and are
   an excellent basis for our own analysis.  The main differences are as
   follows:

   o  In MIPv6 RO, the address binding affects all the communications
      involving an address, while in the MPTCP case, a single connection
      is at stake.  If a binding between two addresses is created at the
      IP layer, this binding can and will affect all the connections
      that involve those addresses.  However, in MPTCP, if an additional
      address is added to an ongoing TCP connection, the additional
      address will/can only affect the connection at hand and not other
      connections, even if the same address is being used for those
      other connections.  The result is that, in MPTCP, there is much
      less at stake and the resulting vulnerabilities are less.  On the
      other hand, it is very important to keep the assumption valid that
      the address bindings for a given connection do not affect other
      connections.  If reusing of binding or security information is to
      be considered, this assumption could be no longer valid and the
      full impact of the vulnerabilities must be assessed.






Bagnulo                       Informational                     [Page 4]

RFC 6181                  MPTCP Threat Analysis               March 2011


   o  In MIPv6, there is a trusted third party, called the Home Agent
      that can help with some security problems, as expanded in the next
      bullet.

   o  In MIPv6 RO, there is the assumption that the original address
      (Home Address) through which the connection has been established
      is always available, and in case it is not, the communication will
      be lost.  This is achieved by leveraging in the on the trusted
      party (the Home Agent) to relay the packets to the current
      location of the Mobile Node.  In MPTCP, it is an explicit goal to
      provide communication resilience when one of the address pairs is
      no longer usable, so it is not possible to leverage on the
      original address pair to be always working.

   o  MIPv6 RO is, of course, designed for IPv6, and it is an explicit
      goal of MPTCP to support both IPv6 and IPv4.  Some MIPv6 RO
      security solutions rely on the usage of some characteristics of
      IPv6 (such as the usage of Cryptographically Generated Addresses
      (CGA) [RFC3972]), which will not be usable in the context of
      MPTCP.

   o  As opposed to MPTCP, MIPv6 RO does not have connection-state-
      information, including sequence numbers, port numbers that could
      be leveraged to provide security in some form.

   In the Shim6 [RFC5533] design, similar issues related to address
   agility were considered and a threat analysis was also performed
   [RFC4218].  The analysis performed for Shim6 also largely applies to
   the MPTCP context, the main differences being:

   o  The Shim6 protocol is a layer 3 protocol so all the communications
      involving the target address are at stake; in MPTCP, the impact
      can be limited to a single TCP connection.

   o  Similar to MIPv6 RO, Shim6 only uses IPv6 addresses as identifiers
      and leverages on some of their properties to provide the security,
      such as relying on CGA or Hash-Based Addresses (HBA) [RFC5535],
      which is not possible in the MPTCP case where IPv4 addresses must
      be supported.

   o  Similar to MIPv6 RO, Shim6 does not have a connection-state-
      information, including sequence numbers, port that could be
      leveraged to provide security in some form.

   Stream Control Transmission Protocol (SCTP) [RFC4960]is a transport
   protocol that supports multiple addresses per endpoint and the
   security implications are very close to the ones of MPTCP.  A
   security analysis, identifying a set of attacks and proposed



Bagnulo                       Informational                     [Page 5]

RFC 6181                  MPTCP Threat Analysis               March 2011


   solutions was performed in [RFC5062].  The results of this analysis
   apply directly to the case of MPTCP.  However, the analysis was
   performed after the base SCTP was designed and the goal of the
   document was essentially to improve the security of SCTP.  As such,
   the document is very specific to the actual SCTP specification and
   relies on the SCTP messages and behavior to characterize the issues.
   While some them can be translated to the MPTCP case, some may be
   caused by the specific behavior of SCTP.

   So, the conclusion is that while there is significant amount of
   previous work that is closely related, and it can and will be used it
   as a basis for this analysis, there is a set of characteristics that
   are specific to MPTCP that grant the need for a specific analysis for
   MPTCP.  The goal of this analysis is to help MPTCP designers to
   include a set of security mechanisms that prevent the introduction of
   new vulnerabilities to the Internet due to the adoption of MPTCP.

4.  Basic MPTCP



   The goal of this document is to serve as input for MPTCP designers to
   properly take into account the security issues.  As such, the
   analysis cannot be performed for a specific MPTCP specification, but
   must be a general analysis that applies to the widest possible set of
   MPTCP designs.  In order to do that, the fundamental features that
   any MPTCP must provide are identified and only those are assumed
   while performing the security analysis.  In some cases, there is a
   design choice that significantly influences the security aspects of
   the resulting protocol.  In that case, both options are considered.

   It is assumed that any MPTCP will behave in the case of a single
   address per endpoint as TCP.  This means that an MPTCP connection
   will be established by using the TCP 3-way handshake and will use a
   single address pair.

   The addresses used for the establishment of the connection do have a
   special role in the sense that this is the address used as identifier
   by the upper layers.  The address used as destination address in the
   SYN packet is the address that the application is using to identify
   the peer and has been obtained either through the DNS (with or
   without DNS Security (DNSSEC) validation) or passed by a referral or
   manually introduced by the user.  As such, the initiator does have a
   certain amount of trust in the fact that it is establishing a
   communication with that particular address.  If due to MPTCP, packets
   end up being delivered to an alternative address, the trust that the
   initiator has placed on that address would be deceived.  In any case,
   the adoption of MPTCP necessitates a slight evolution of the
   traditional TCP trust model, in that the initiator is additionally
   trusting the peer to provide additional addresses that it will trust



Bagnulo                       Informational                     [Page 6]

RFC 6181                  MPTCP Threat Analysis               March 2011


   to the same degree as the original pair.  An application or
   implementation that cannot trust the peer in this way should not make
   use of multiple paths.

   During the 3-way handshake, the sequence number will be synchronized
   for both ends, as in regular TCP.  It is assumed that an MPTCP
   connection will use a single sequence number for the data, even if
   the data is exchanged through different paths, as MPTCP provides an
   in-order delivery service of bytes

   Once the connection is established, the MPTCP extensions can be used
   to add addresses for each of the endpoints.  This is achieved by each
   end sending a control message containing the additional address(es).
   In order to associate the additional address to an ongoing
   connection, the connection needs to be identified.  It is assumed
   that the connection can be identified by the 4-tuple of source
   address, source port, destination address, destination port used for
   the establishment of the connection.  So, at least, the control
   message that will convey the additional address information can also
   contain the 4-tuple in order to inform about what connection the
   address belong to (if no other connection identifier is defined).
   There are two different ways to convey address information:

   o  Explicit mode: the control message contain a list of addresses.

   o  Implicit mode: the address added is the one included in the source
      address field of the IP header

   These two modes have different security properties for some type of
   attacks.  The explicit mode seems to be the more vulnerable to abuse.
   The implicit mode may benefit from forms of ingress filtering
   security, which would reduce the possibility of an attacker to add
   any arbitrary address to an ongoing connection.  However, ingress
   filtering deployment is far from universal, and it is unwise to rely
   on it as a basis for the protection of MPTCP.

   Further consideration regarding the interaction between ingress
   filtering and implicit mode signaling is needed in the case that an
   address that is no longer available from the MPTCP connection is
   removed.  A host attached to a network that performs ingress
   filtering and using implicit signaling would not be able to remove an
   address that is no longer available (either because of a failure or
   due to a mobility event) from an ongoing MPTCP connection.

   It is assumed that MPTCP will use all the address pairs that it has
   available for sending packets, and that it will distribute the load
   based on congestion among the different paths.




Bagnulo                       Informational                     [Page 7]

RFC 6181                  MPTCP Threat Analysis               March 2011


5.  Flooding Attacks



   The first type of attacks that are introduced by address agility are
   the flooding (or bombing) attacks.  The setup for this attack is
   depicted in the following figure:

               +--------+        (step 1)           +------+
               |Attacker| ------------------------- |Source|
               |    A   |IPA                     IPS|  S   |
               +--------+                          /+------+
                                                  /
                                        (step 2) /
                                                /
                                               v IPT
                                           +------+
                                           |Target|
                                           |  T   |
                                           +------+

   The scenario consists of an Attacker A who has an IP address IPA.  A
   server that can generate a significant amount of traffic (such as a
   streaming server), called source S and that has IP address IPS.
   Target T has an IP address IPT.

   In step 1 of this attack, the Attacker A establishes an MPTCP
   connection with the source of the traffic server S and starts
   downloading a significant amount of traffic.  The initial connection
   only involves one IP address per endpoint, IPA and IPS.  Once the
   download is on course, in step 2 of the attack, the Attacker A adds
   IPT as one of the available addresses for the communication.  How the
   additional address is added depends on the MPTCP address management
   mode.  In explicit address management, the Attacker A only needs to
   send a signaling packet conveying address IPT.  In implicit mode, the
   Attacker A would need to send a packet with IPT as the source
   address.  Depending on whether ingress filtering is deployed and the
   location of the attacker, it may or may not be possible for the
   attacker to send such a packet.  At this stage, the MPTCP connection
   still has a single address for the Source S, i.e., IPS, but has two
   addresses for the Attacker A, IPA, and IPT.  The attacker now
   attempts to get the Source S to send the traffic of the ongoing
   download to the Target T IP address, i.e., IPT.  The attacker can do
   that by pretending that the path between IPA and IPT is congested but
   that the path between IPS and IPT is not.  In order to do that, it
   needs to send ACKs for the data that flows through the path between
   IPS and IPT and not send ACKs for the data that is sent to IPA.  The
   details of this will depend on how the data sent through the
   different paths is ACKed.  One possibility is that ACKs for the data
   sent using a given address pair should come in packets containing the



Bagnulo                       Informational                     [Page 8]

RFC 6181                  MPTCP Threat Analysis               March 2011


   same address pair.  If so, the attacker would need to send ACKs using
   packets containing IPT as the source address to keep the attack
   flowing.  This may or may not be possible depending on the deployment
   of ingress filtering and the location of the attacker.  The attacker
   would also need to guess the sequence number of the data being sent
   to the Target.  Once the attacker manages to perform these actions,
   the attack is on place and the download will hit the target.  In this
   type of attack, the Source S still thinks it is sending packets to
   the Attacker A while in reality it is sending the packet to Target T.

   Once the traffic from the Source S start hitting the Target T, the
   target will react.  Since the packets are likely to belong to a non-
   existent TCP connection, the Target T will issue RST packets.  It is
   relevant to understand how MPTCP reacts to incoming RST packets.  It
   seems that the at least the MPTCP that receives a RST packet should
   terminate the packet exchange corresponding to the particular address
   pair (maybe not the complete MPTCP connection, but at least it should
   not send more packets with the address pair involved in the RST
   packet).  However, if the attacker, before redirecting the traffic
   has managed to increase the window size considerably, the flight size
   could be enough to impose a significant amount of traffic to the
   Target node.  There is a subtle operation that the attacker needs to
   achieve in order to launch a significant attack.  On the one hand, it
   needs to grow the window enough so that the flight size is big enough
   to cause enough effect; on the other hand, the attacker needs to be
   able to simulate congestion on the IPA-IPS path so that traffic is
   actually redirected to the alternative path without significantly
   reducing the window.  This will heavily depend on how the coupling of
   the windows between the different paths works, in particular how the
   windows are increased.  Some designs of the congestion control window
   coupling could render this attack ineffective.  If the MPTCP requires
   performing slow start per subflow, then the flooding will be limited
   by the slow-start initial window size.

   Previous protocols, such as MIPv6 RO and SCTP, that have to deal with
   this type of attacks have done so by adding a reachability check
   before actually sending data to a new address.  The solution used in
   other protocols would include the Source S to explicitly asking the
   host sitting in the new address (the Target T sitting in IPT) whether
   it is willing to accept packets from the MPTCP connection identified
   by the 4-tuple IPA, port A, IPS, port S.  Since this is not part of
   the established connection that Target T has, T would not accept the
   request and Source S would not use IPT to send packets for this MPTCP
   connection.  Usually, the request also includes a nonce that cannot
   be guessed by the Attacker A so that it cannot fake the reply to the
   request easily.  In the case of SCTP, it sends a message with a 64-
   bit nonce (in a HEARTBEAT).




Bagnulo                       Informational                     [Page 9]

RFC 6181                  MPTCP Threat Analysis               March 2011


   One possible approach to do this reachability test would be to
   perform a 3-way handshake for each new address pair that is going to
   be used in an MPTCP connection.  While there are other reasons for
   doing this (such as NAT traversal), such approach would also act as a
   reachability test and would prevent the flooding attacks described in
   this section.

   Another type of flooding attack that could potentially be performed
   with MPTCP is one where the attacker initiates a communication with a
   peer and includes a long list of alternative addresses in explicit
   mode.  If the peer decides to establish subflows with all the
   available addresses, the attacker has managed to achieve an amplified
   attack, since by sending a single packet containing all the
   alternative addresses, it triggers the peer to generate packets to
   all the destinations.

6.  Hijacking Attacks



6.1.  Hijacking Attacks to the Basic MPTCP



   The hijacking attacks essentially use the MPTCP address agility to
   allow an attacker to hijack a connection.  This means that the victim
   of a connection thinks that it is talking to a peer, while it is
   actually exchanging packets with the attacker.  In some sense, it is
   the dual of the flooding attacks (where the victim thinks it is
   exchanging packets with the attacker but in reality is sending the
   packets to the target).

   The scenario for a hijacking attack is described in the next figure.

                +------+                           +------+
                | Node | ------------------------- | Node |
                |   1  |IP1                     IP2|  2   |
                +------+                          /+------+
                                                 /
                                                /
                                               /
                                              v IPA
                                         +--------+
                                         |Attacker|
                                         |    A   |
                                         +--------+

   An MPTCP connection is established between Node 1 and Node 2.  The
   connection is using only one address per endpoint, IP1 and IP2.  The
   attacker then launches the hijacking attack by adding IPA as an
   additional address for Node 1.  There is not much difference between
   explicit or implicit address management, since, in both cases, the



Bagnulo                       Informational                    [Page 10]

RFC 6181                  MPTCP Threat Analysis               March 2011


   Attacker A could easily send a control packet adding the address IPA,
   either as control data or as the source address of the control
   packet.  In order to be able to hijack the connection, the attacker
   needs to know the 4-tuple that identifies the connection, including
   the pair of addresses and the pair of ports.  It seems reasonable to
   assume that knowing the source and destination IP addresses and the
   port of the server side is fairly easy for the attacker.  Learning
   the port of the client (i.e., of the initiator of the connection) may
   prove to be more challenging.  The attacker would need to guess what
   the port is or to learn it by intercepting the packets.  Assuming
   that the attacker can gather the 4-tuple and issue the message adding
   IPA to the addresses available for the MPTCP connection, then the
   Attacker A has been able to participate in the communication.  In
   particular:

   o  Segments flowing from the Node 2: Depending how the usage of
      addresses is defined, Node 2 will start using IPA to send data to.
      In general, since the main goal is to achieve multipath
      capabilities, it can be assumed that unless there are already many
      IP address pairs in use in the MPTCP connection, Node 2 will start
      sending data to IPA.  This means that part of the data of the
      communication will reach the attacker but probably not all of it.
      This already has negative effects, since Node 1 will not receive
      all the data from Node 2.  Moreover, from the application
      perspective, this would result in a Denial-of-Service (DoS)
      attack, since the byte flow will stop waiting for the missing
      data.  However, it is not enough to achieve full hijacking of the
      connection, since part of data will be still delivered to IP1, so
      it would reach Node 1 and not the attacker.  In order for the
      attacker to receive all the data of the MPTCP connection, the
      attacker must somehow remove IP1 of the set of available addresses
      for the connection.  In the case of implicit address management,
      this operation is likely to imply sending a termination packet
      with IP1 as source address, which may or may not be possible for
      the attacker depending on whether ingress filtering is in place
      and the location of the attacker.  If explicit address management
      is used, then the attacker will send a remove address control
      packet containing IP1.  Once IP1 is removed, all the data sent by
      Node 2 will reach the attacker and the incoming traffic has been
      hijacked.











Bagnulo                       Informational                    [Page 11]

RFC 6181                  MPTCP Threat Analysis               March 2011


   o  Segments flowing to the Node 2: As soon as IPA is accepted by Node
      2 as part of the address set for the MPTCP connection, the
      attacker can send packets using IPA, and those packets will be
      considered as part of MPTCP connection by Node 2.  This means that
      the attacker will be able to inject data into the MPTCP
      connection, so from this perspective, the attacker has hijacked
      part of the outgoing traffic.  However, Node 1 would still be able
      to send traffic that will be received by Node 2 as part of the
      MPTCP connection.  This means that there will be two sources of
      data, i.e., Node 1 and the attacker, potentially preventing the
      full hijacking of the outgoing traffic by the attacker.  In order
      to achieve a full hijacking, the attacker would need to remove IP1
      from the set of available addresses.  This can be done using the
      same techniques described in the previous paragraph.

   A related attack that can be achieved using similar techniques would
   be an MiTM attack.  The scenario for the attack is depicted in the
   figure below.

                        +------+                 +------+
                        | Node | --------------- | Node |
                        |   1  |IP1           IP2|  2   |
                        +------+ \              /+------+
                                  \            /
                                   \          /
                                    \        /
                                    v IPA  v
                                   +--------+
                                   |Attacker|
                                   |    A   |
                                   +--------+

   There is an established connection between Node 1 and Node 2.  The
   Attacker A will use the MPTCP address agility capabilities to place
   itself as a MiTM.  In order to do so, it will add IP address IPA as
   an additional address for the MPTCP connection on both Node 1 and
   Node 2.  This is essentially the same technique described earlier in
   this section, only that it is used against both nodes involved in the
   communication.  The main difference is that in this case, the
   attacker can simply sniff the content of the communication that is
   forwarded through it and in turn forward the data to the peer of the
   communication.  The result is that the attacker can place himself in
   the middle of the communication and sniff part of the traffic
   unnoticed.  Similar considerations about how the attacker can manage
   to get to see all the traffic by removing the genuine address of the
   peer apply.





Bagnulo                       Informational                    [Page 12]

RFC 6181                  MPTCP Threat Analysis               March 2011


6.2.  Time-Shifted Hijacking Attacks



   A simple way to prevent off-path attackers from launching hijacking
   attacks is to provide security for the control messages that adds and
   removes addresses by the usage of a cookie.  In this type of
   approaches, the peers involved in the MPTCP connection agree on a
   cookie that is exchanged in plaintext during the establishment of the
   connection and that needs to be presented in every control packet
   that adds or removes an address for any of the peers.  The result is
   that the attacker needs to know the cookie in order to launch any of
   the hijacking attacks described earlier.  This implies that off-path
   attackers can no longer perform the hijacking attacks and that only
   on-path attackers can do so, so one may consider a cookie-based
   approach to secure MPTCP connection results in similar security to
   current TCP.  While it is close, it is not entirely true.

   The main difference between the security of an MPTCP secured through
   cookies and the current TCP is the time-shifted attacks.  As has been
   described earlier, a time-shifted attack is one where the attacker is
   along the path during a period of time, and then moves away but the
   effects of the attack still remain, after the attacker is long gone.
   In the case of an MPTCP secured through the usage of cookies, the
   attacker needs to be along the path until the cookie is exchanged.
   After the attacker has learned the cookie, it can move away from the
   path and can still launch the hijacking attacks described in the
   previous section.

   There are several types of approaches that provide some protection
   against hijacking attacks and that are vulnerable to some forms of
   time-shifted attacks.  A general taxonomy of solutions and the
   residual threats for each type is presented next:

   o  Cookie-based solution: As it has been described earlier, one
      possible approach is to use a cookie that is sent in cleartext in
      every MPTCP control message that adds a new address to the
      existing connection.  The residual threat in this type of solution
      is that any attacker that can sniff any of these control messages
      will learn the cookie and will be able to add new addresses at any
      given point in the lifetime of the connection.  Moreover, the
      endpoints will not detect the attack since the original cookie is
      being used by the attacker.  Summarizing, the vulnerability window
      of this type of attacks includes all the flow establishment
      exchanges and it is undetectable by the endpoints.

   o  Shared secret exchanged in plaintext: An alternative option that
      is more secure than the cookie-based approach is to exchange a key
      in cleartext during the establishment of the first subflow and
      then validate the following subflows by using a keyed Hashed



Bagnulo                       Informational                    [Page 13]

RFC 6181                  MPTCP Threat Analysis               March 2011


      Message Authentication Code (HMAC) signature using the shared key.
      This solution would be vulnerable to attackers sniffing the
      message exchange for the establishment of the first subflow, but
      after that, the shared key is not transmitted any more, so the
      attacker cannot learn it through sniffing any other message.
      Unfortunately, in order to be compatible with NATs (see analysis
      below) even though this approach includes a keyed HMAC signature,
      this signature cannot cover the IP address that is being added.
      This means that this type of approaches are also vulnerable to
      integrity attacks of the exchanged messages.  This means that even
      though the attacker cannot learn the shared key by sniffing the
      subsequent subflow establishment, the attacker can modify the
      subflow establishment message and change the address that is being
      added.  So, the vulnerability window for confidentially to the
      shared key is limited to the establishment of the first subflow,
      but the vulnerability window for integrity attacks still includes
      all the subflow establishment exchanges.  These attacks are still
      undetectable by the endpoints.  The SCTP security falls in this
      category.

   o  Strong crypto anchor exchange: Another approach that could be used
      would be to exchange some strong crypto anchor while the
      establishment of the first subflow, such as a public key or a hash
      chain anchor.  Subsequent subflows could be protected by using the
      crypto material associated to that anchor.  An attacker in this
      case would need to change the crypto material exchanged in the
      connection establishment phase.  As a result, the vulnerability
      window for forging the crypto anchor is limited to the initial
      connection establishment exchange.  Similar to the previous case,
      due to NAT traversal considerations, the vulnerability window for
      integrity attacks include all the subflow establishment exchanges.
      Because the attacker needs to change the crypto anchor, this
      approach are detectable by the endpoints, if they communicate
      directly.

6.3.  NAT Considerations



   In order to be widely adopted, MPTCP must work through NATs.  NATs
   are an interesting device from a security perspective.  In terms of
   MPTCP, they essentially behave as an MiTM attacker.  MPTCP's security
   goal is to prevent from any attacker to insert their addresses as
   valid addresses for a given MPTCP connection.  But that is exactly
   what a NAT does: it modifies the addresses.  So, if MPTCP is to work
   through NATs, MPTCP must accept address rewritten by NATs as valid
   addresses for a given session.  The most direct corollary is that the
   MPTCP messages that add addresses in the implicit mode (i.e., the SYN
   of new subflows) cannot be protected against integrity attacks, since
   they must allow for NATs to change their addresses.  This rules out



Bagnulo                       Informational                    [Page 14]

RFC 6181                  MPTCP Threat Analysis               March 2011


   any solution that would rely on providing integrity protection to
   prevent an attacker from changing the address used in a subflow
   establishment exchange.  This implies that alternative creative
   mechanisms are needed to protect from integrity attacks to the MPTCP
   signaling that adds new addresses to a connection.  It is far from
   obvious how one such creative approach could look like at this point.

   In the case of explicit mode, you could protect the address included
   in the MPTCP option.  Now the question is what address to include in
   the MPTCP option that conveys address information.  If the address
   included is the address configured in the host interface and that
   interface is behind a NAT, the address information is useless, as the
   address is not actually reachable from the other end so there is no
   point in conveying it and even less in securing it.  It would be
   possible to envision the usage of NAT traversal techniques, such as
   Session Traversal Utilities for NAT (STUN) to learn the address and
   port that the NAT has assigned and convey that information in a
   secure.  While this is possible, it relies on using NAT traversal
   techniques and also tools to convey the address and the port in a
   secure manner.

7.  Recommendation



   The presented analysis shows that there is a tradeoff between the
   complexity of the security solution and the residual threats.  After
   evaluating the different aspects in the MPTCP WG, the conclusions are
   as follows:

   MPTCP should implement some form of reachability check using a random
   nonce (e.g., TCP 3-way handshake) before adding a new address to an
   ongoing communication in order to prevent flooding attacks.

   The default security mechanisms for MPTCP should be to exchange a key
   in cleartext in the establishment of the first subflow and then
   secure following address additions by using a keyed HMAC using the
   exchanged key.

   MPTCP security mechanism should support using a pre-shared key to be
   used in the keyed HMAC, providing a higher level of protection than
   the previous one.

   A mechanism to prevent replay attacks using these messages should be
   provided, e.g., a sequence number protected by the HMAC.

   The MPTCP should be extensible and it should be able to accommodate
   multiple security solutions, in order to enable the usage of more
   secure mechanisms if needed.




Bagnulo                       Informational                    [Page 15]

RFC 6181                  MPTCP Threat Analysis               March 2011


8.  Security Considerations



   This note contains a security analysis for MPTCP, so no further
   security considerations need to be described in this section.

9.  Contributors



   Alan Ford - Roke Manor Research, Ltd.

10.  Acknowledgments



   Rolf Winter, Randall Stewart, Andrew McDonald, Michael Tuexen,
   Michael Scharf, Tim Shepard, Yoshifumi Nishida, Lars Eggert, Phil
   Eardley, Jari Arkko, David Harrington, Dan Romascanu, and Alexey
   Melnikov reviewed an earlier version of this document and provided
   comments to improve it.

   Mark Handley pointed out the problem with NATs and integrity
   protection of MPTCP signaling.

   Marcelo Bagnulo is partly funded by Trilogy, a research project
   supported by the European Commission under its Seventh Framework
   Program.

11.  References



11.1.  Normative References



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

11.2.  Informative References



   [RFC4225]  Nikander, P., Arkko, J., Aura, T., Montenegro, G., and E.
              Nordmark, "Mobile IP Version 6 Route Optimization Security
              Design Background", RFC 4225, December 2005.

   [RFC4218]  Nordmark, E. and T. Li, "Threats Relating to IPv6
              Multihoming Solutions", RFC 4218, October 2005.

   [RFC3972]  Aura, T., "Cryptographically Generated Addresses (CGA)",
              RFC 3972, March 2005.

   [RFC5062]  Stewart, R., Tuexen, M., and G. Camarillo, "Security
              Attacks Found Against the Stream Control Transmission
              Protocol (SCTP) and Current Countermeasures", RFC 5062,
              September 2007.




Bagnulo                       Informational                    [Page 16]

RFC 6181                  MPTCP Threat Analysis               March 2011


   [RFC5535]  Bagnulo, M., "Hash-Based Addresses (HBA)", RFC 5535,
              June 2009.

   [RFC3775]  Johnson, D., Perkins, C., and J. Arkko, "Mobility Support
              in IPv6", RFC 3775, June 2004.

   [RFC5533]  Nordmark, E. and M. Bagnulo, "Shim6: Level 3 Multihoming
              Shim Protocol for IPv6", RFC 5533, June 2009.

   [RFC4960]  Stewart, R., "Stream Control Transmission Protocol",
              RFC 4960, September 2007.

   [MPTCP-MULTIADDRESSED]
              Ford, A., Raiciu, C., and M. Handley, "TCP Extensions for
              Multipath Operation with Multiple Addresses", Work
              in Progress, October 2010.

Author's Address



   Marcelo Bagnulo
   Universidad Carlos III de Madrid
   Av. Universidad 30
   Leganes, Madrid  28911
   SPAIN

   Phone: 34 91 6248814
   EMail: marcelo@it.uc3m.es
   URI:   http://www.it.uc3m.es























Bagnulo                       Informational                    [Page 17]