wdiff draft-ietf-ipsecme-ikev2-fragmentation-07.txt draft-ietf-ipsecme-ikev2-fragmentation-08.txt

Network Working Group V. Smyslov
Internet-Draft ELVIS-PLUS
Intended status: Standards Track April 4, May 23, 2014
Expires: October 6, November 24, 2014

IKEv2 Fragmentation
draft-ietf-ipsecme-ikev2-fragmentation-07
draft-ietf-ipsecme-ikev2-fragmentation-08

Abstract

This document describes the way to avoid IP fragmentation of large
IKEv2 messages. This allows IKEv2 messages to traverse network
devices that don't do not allow IP fragments to pass through.

Status of this Memo

This Internet-Draft is submitted in full conformance with the
provisions of BCP 78 and BCP 79.

Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF). Note that other groups may also distribute
working documents as Internet-Drafts. The list of current Internet-
Drafts is at http://datatracker.ietf.org/drafts/current/.

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."

This Internet-Draft will expire on October 6, November 24, 2014.

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
1.1. Problem description . . . . . . . . . . . . . . . . . . . 3
1.2. Proposed solution . . . . . . . . . . . . . . . . . . . . 3
1.3. Conventions Used in This Document . . . . . . . . . . . . 3 4
2. Protocol details . . . . . . . . . . . . . . . . . . . . . . . 4 5
2.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . 4 5
2.2. Limitations . . . . . . . . . . . . . . . . . . . . . . . 4 5
2.3. Negotiation . . . . . . . . . . . . . . . . . . . . . . . 4 5
2.4. Using IKE Fragmentation . . . . . . . . . . . . . . . . . 5 6
2.5. Fragmenting Message . . . . . . . . . . . . . . . . . . . 6 7
2.5.1. Selecting Fragment Size . . . . . . . . . . . . . . . 8 9
2.5.2. PMTU Discovery . . . . . . . . . . . . . . . . . . . . 8 10
2.5.3. Fragmenting Messages containing unencrypted unprotected
Payloads . . . . . . . . . . . . . . . . . . . . . . . 10 11
2.6. Receiving IKE Fragment Message . . . . . . . . . . . . . . 10 12
2.6.1. Changes in Replay Protection Logic . Detection and Retransmissions . . . . . . . . . 12 14
3. Interaction with other IKE extensions . . . . . . . . . . . . 13 15
4. Transport Considerations . . . . . . . . . . . . . . . . . . . 14 16
5. Security Considerations . . . . . . . . . . . . . . . . . . . 15 17
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 18
7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 17 19
8. References . . . . . . . . . . . . . . . . . . . . . . . . . . 18 20
8.1. Normative References . . . . . . . . . . . . . . . . . . . 18 20
8.2. Informative References . . . . . . . . . . . . . . . . . . 18 20
Appendix A. Design rationale . . . . . . . . . . . . . . . . . . 20 22
Appendix B. Correlation between IP Datagram size and
Encrypted Payload content size . . . . . . . . . . . 21 23
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 23 25

1. Introduction

1.1. Problem description

The Internet Key Exchange Protocol version 2 (IKEv2), specified in
[RFC5996],
[IKEv2], uses UDP as a transport for its messages. Most IKEv2
messages are relatively small, usually below several hundred bytes.
Noticeable exception is IKE_AUTH exchange, Exchange, which requires fairly
large messages, up to several kbytes, especially when certificates
are transferred. When IKE message size exceeds path MTU, it gets
fragmented by IP level. The problem is that some network devices,
specifically some NAT boxes, don't do not allow IP fragments to pass
through. This apparently blocks IKE communication and, therefore,
prevents peers from establishing IPsec SA. Section 2 of [IKEv2]
discusses the impact of IP fragmentation on IKEv2 and acknowledges
this problem.

Widespread deployment of Carrier-Grade NATs (CGN) introduces new
challenges. RFC6888 [RFC6888] describes requirements for CGNs. It states,
that CGNs must comply with Section 11 of RFC4787 [RFC4787], which requires
NAT to support receiving IP fragments (REQ-14). In real life
fulfillment of this requirement creates an additional burden in terms
of memory, especially for high-capacity devices, used in CGNs. It
was found by people deploying IKE, that some more and more ISPs have
begun to use
equipment that drop IP fragments, violating that this requirement.

Security researchers have found and continue to find attack vectors
that rely on IP fragmentation. For these reasons, and also
articulated in [FRAGDROP], many network operators filter all IPv6
fragments. Also, the default behavior of many currently deployed
firewalls is to discard IPv6 fragments.

In one recent study [BLACKHOLES], two researchers utilized a
measurement network to measure fragment filtering. They sent
packets, fragmented to the minimum MTU of 1280, to 502 IPv6 enabled
and reachable probes. They found that during any given trial period,
ten percent of the probes did not receive fragmented packets.

Thus this problem is valid for both IPv4 and IPv6 and may be caused
either by deficiency of network devices or by operational choice.

1.2. Proposed solution

The solution to the problem described in this document is to perform
fragmentation of large messages by IKE IKEv2 itself, replacing them by
series of smaller messages. In this case the resulting IP Datagrams
will be small enough so that no fragmentation on IP level will take
place.

Avoiding IP fragmentation

The primary goal of this solution is beneficial for to allow IKEv2 to operate in general.
Security Considerations Section of [RFC5996] mentions exhausting of
the
environments, that may block IP reassembly buffers as one of possible attacks on the protocol.
In the paper [DOSUDPPROT] several aspects of attacks on IKE using fragments. This goal does not assume
that IP fragmentation are discussed, and one of should be avoided completely, but only in those
cases when it interferes with IKE operations. However this solution
could be used to avoid IP fragmentation in all situations where
fragmentation within IKE is applicable, as it is recommended in
Section 3.2 of [RFC5405]. Avoiding IP fragmentation would be
beneficial for IKEv2 in general. Security Considerations Section of
[IKEv2] mentions exhausting of the IP reassembly buffers as one of
the possible attacks on the protocol. In the paper [DOSUDPPROT]
several aspects of attacks on IKE using IP fragmentation are
discussed, and one of the defenses it proposes is to perform IKE-level fragmentation, similar
fragmentation within IKE similarly to the solution, solution described in this
document.

1.1.

1.3. Conventions Used in This Document

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. Protocol details

2.1. Overview

The idea of the protocol is to split large IKE IKEv2 message into a set
of smaller ones, called IKE Fragment Messages. Fragmentation takes
place before the original message is encrypted and authenticated, so
that each IKE Fragment Message receives individual protection. On
the receiving side IKE Fragment Messages are collected, verified,
decrypted and merged together to get the original message before
encryption. For design rationale see See Appendix A. A for design rationale.

2.2. Limitations

Since IKE Fragment Messages are cryptographically protected, SK_a and
SK_e must already be calculated. In general, it means that original
message can be fragmented if and only if it contains Encrypted
Payload.

This implies that messages of the IKE_SA_INIT Exchange cannot be
fragmented. In most cases this is not a problem, since problem because IKE_SA_INIT
messages are usually small enough to avoid IP fragmentation. But in
some cases (advertising a badly structured long list of algorithms,
using large MODP Groups, etc.) these messages may become fairly large
and get fragmented by IP level. In this case the described solution
won't
will not help.

Among existing IKEv2 extensions, messages of IKE_SESSION_RESUME
Exchange, defined in [RFC5723], cannot be fragmented either. See
Section 3 for details.

Another limitation is that the minimal size of IP Datagram bearing
IKE Fragment Message is about 100 bytes depending on the algorithms
employed. According to [RFC0791] the minimum IP IPv4 Datagram size that
is guaranteed not to be further fragmented is 68 bytes. So, even the
smallest IKE Fragment Messages could be fragmented by IP level in
some circumstances. But such extremely small PMTU sizes are very
rare in real life.

2.3. Negotiation

Initiator MAY indicate indicates its support for the IKE Fragmentation and
willingness to use it by including Notification Payload of type
IKEV2_FRAGMENTATION_SUPPORTED in IKE_SA_INIT request message. If
Responder also supports this extension and is willing to use it, it
includes this notification in response message.

Initiator Responder
----------- -----------
HDR, SAi1, KEi, Ni,
[N(IKEV2_FRAGMENTATION_SUPPORTED)] -->

<-- HDR, SAr1, KEr, Nr, [CERTREQ],
[N(IKEV2_FRAGMENTATION_SUPPORTED)]

The Notify payload is formatted as follows:

o Protocol ID (1 octet) MUST be 0.

o SPI Size (1 octet) MUST be 0, meaning no SPI is present.

o Notify Message Type (2 octets) - MUST be xxxxx, the value assigned
for IKEV2_FRAGMENTATION_SUPPORTED by IANA. notification.

This Notification contains no data.

2.4. Using IKE Fragmentation

The IKE Fragmentation MUST NOT be used unless both peers have
indicated their support for it. After IKE Fragmentation is negotiated, that it is up to the the
Initiator of each Exchange, exchange to decide whether to use it or not. In most cases
IKE Fragmentation will be used The
Responder usually replies in IKE_AUTH Exchange, especially if
certificates are employed. the same form as the request message,
but other considerations might override this.

The Initiator may employ various policies regarding the use of IKE
Fragmentation. It may first try to send an unfragmented message and
resend it as fragmented only if it didn't
receive no complete response is received even
after several retransmissions, or retransmissions. Alternatively, it may choose always
to send
messages fragmented messages (but see Section 3), or it may fragment
only large messages and messages causing that are expected to result in large
responses.

In general the

The following general guidelines are applicable for initiator: apply:

o Initiator MAY fragment outgoing message if it If either peer has some knowledge
(possibly from lower layer or from configuration) or suspicions information that either request or response message will a part of the transaction is
likely to be fragmented by at the IP
level.

o Initiator layer, causing interference with
the IKE exchange, that peer SHOULD fragment outgoing message if it has some
knowledge (possibly use IKE Fragmentation. This
information may be passed from a lower layer or from configuration) or
suspicions that either request layer, provided by
configuration, or derived through heuristics. Examples of
heuristics are the lack of a complete response message will be
fragmented by IP level after several
retransmissions for the Initiator, and receiving repeated
retransmissions of the request for the Responder.

o If either peer knows that IKE Fragmentation was already has been used in
one of a
previous Exchanges exchange in the context of the current IKE SA. SA, that peer
SHOULD continue the use of IKE Fragmentation for the messages that
are larger than the current fragmentation threshold (see
Section 2.5.1).

o Initiator IKE Fragmentation SHOULD NOT fragment outgoing message if be used in cases where IP-layer
fragmentation of both the request and response messages of the Exchange are small enough not to cause
fragmentation on IP level (for is
unlikely. For example, there is no point in fragmenting Liveness
Check messages).

In general the following guidelines are applicable for responder: messages.

o If none of the above apply, the Responder SHOULD send response message respond in the
same form (fragmented or not) as corresponded request message. If it
received unfragmented request message, responded with unfragmented
response message and then receives fragmented retransmission of the same request, it SHOULD resend its response back to Initiator
fragmented.

o Responder MAY respond to unfragmented request message with fragmented
response if it has some knowledge (possibly from lower layer or
from configuration) or suspicions is
responding to. Note that response message will be
fragmented by IP level.

o Responder MAY respond to fragmented message with unfragmented
response if the size other guidelines might override this
because of information or heuristics available to the response message is less than the
smallest fragmentation threshold, supported by Responder (for
example, there is no point Responder.

In most cases IKE Fragmentation will be used in fragmenting Liveness Check
messages). the IKE_AUTH
Exchange, especially if certificates are employed.

2.5. Fragmenting Message

Message to be fragmented MUST contain Encrypted Payload. For the
purpose of IKE Fragment Messages construction original (unencrypted)
content of Encrypted Payload is split into chunks. The content is
treated as a binary blob and is split regardless of inner Payloads
boundaries. Each of resulting chunks is treated as an original
content of Encrypted Fragment Payload and is then encrypted and
authenticated. Thus, the Encrypted Fragment Payload contains a chunk
of the original content of Encrypted Payload in encrypted form. The
cryptographic processing of Encrypted Fragment Payload is identical
to Section 3.14 of [RFC5996], [IKEv2], as well as documents updating it for
particular algorithms or modes, such as [RFC5282].

The Encrypted Fragment Payload, similarly to the Encrypted Payload,
if present in a message, MUST be the last payload in the message.

The Encrypted Fragment Payload is denoted SKF{...} and its payload
type is XXX (TBA by IANA). This payload is also called the
"Encrypted and Authenticated Fragment" payload.

1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Next Payload |C| RESERVED | Payload Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Fragment Number | Total Fragments |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Initialization Vector |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
~ Encrypted content ~
+ +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| | Padding (0-255 octets) |
+-+-+-+-+-+-+-+-+ +-+-+-+-+-+-+-+-+
| | Pad Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
~ Integrity Checksum Data ~
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Encrypted Fragment Payload

o Next Payload (1 octet) - in the very first fragment (with Fragment
Number equal to 1) this field MUST be set to Payload Type of the
first inner Payload (similarly to the Encrypted Payload). In the
rest fragments MUST be set to zero.

o Fragment Number (2 octets) - current fragment number starting from
1. This field MUST be less than or equal to the next field, Total
Fragments. This field MUST NOT be zero.

o Total Fragments (2 octets) - number of fragments original message
was divided into. This field MUST NOT be zero. With PMTU
discovery this field plays additional role. See Section 2.5.2 for
details. This field MUST NOT be
zero.

The other fields are identical to those specified in Section 3.14 of
[RFC5996].
[IKEv2].

When prepending IKE Header, Length field Header to the IKE Fragment Messages it MUST be
taken intact from the original message, except for the Length and the
Next Payload fields. The Length field is adjusted to reflect the
length of the constructed message and the Next Payload field MUST reflect is set
to the payload type of the first Payload in the constructed message (that in (in
most cases it will be Encrypted Fragment Payload). All newly
constructed messages MUST retain the same Message ID as original
message. After prepending
IKE Header and possibly any of all Payloads that precedes possibly precede Encrypted Payload
in original message (see (if any, see Section 2.5.3), the resulting
messages are sent to the peer.

Below is an example of fragmenting a message.

HDR(MID=n), SK(NextPld=PLD1) {PLD1 ... PLDN}

Original Message

HDR(MID=n), SKF(NextPld=PLD1, Frag#=1, TotalFrags=m) {...},
HDR(MID=n), SKF(NextPld=0, Frag#=2, TotalFrags=m) {...},
...
HDR(MID=n), SKF(NextPld=0, Frag#=m, TotalFrags=m) {...}

IKE Fragment Messages

2.5.1. Selecting Fragment Size

When splitting content of Encrypted Payload into chunks sender SHOULD chose
choose their size of those chunks so, that resulting IP Datagram size not exceed Datagrams be smaller than
some fragmentation threshold. Implementation may calculate
fragmentation threshold - be small enough to avoid IP
fragmentation. using various sources of information.

If sender has some knowledge information about PMTU size it MAY SHOULD use it. If
sender is a The
Responder in the Exchange and it has received fragmented
request, it MAY exchange may use maximum size of received IKE
Fragment Message IP Datagrams as threshold when constructing
fragmented response. Successful completion of previous exchanges
(including those exchanges, that cannot employ IKE Fragmentation,
e.g. IKE_SA_INIT) may be an indication, that fragmentation threshold
can be set to the size of the largest of already sent messages.

Otherwise for messages to be sent over IPv6 it is RECOMMENDED to use
value 1280 bytes as a maximum IP Datagram size ([RFC2460]). For
messages to be sent over IPv4 it is RECOMMENDED to use value 576
bytes as a maximum IP Datagram size. Presence of tunnels on the path
may reduce these values. Implementation may use other values if they
are appropriate in current environment.

According to [RFC0791] the minimum IPv4 datagram size that is
guaranteed not to be further fragmented is 68 bytes, but it is
generally impossible to use such small value for solution, described
in this document. Using 576 bytes is a compromise - the value is
large enough for the presented solution and small enough to avoid IP
fragmentation in most situations. Several other UDP-based protocol
assume the value 576 bytes as a safe low limit for IP datagrams size
(Syslog, DNS, etc.). Sender MAY use other values if they are
appropriate.

See Appendix B for correlation between IP Datagram size and Encrypted
Payload content size.

2.5.2. PMTU Discovery

Initiator MAY try to discover path MTU by probing several values

The amount of
fragmentation threshold. While doing probes, node MUST start from
larger values and refragment message with next smaller value if traffic that IKE endpoint produces during lifetime of
IKE SA is fairly modest - usually it
doesn't receive response in is below one hundred kBytes
within a reasonable time after period of several
retransmissions. This time is supposed to be hours. Most of this traffic consists of
relatively short, so
that node could make all desired probes before exchange times out.
When starting new probe (with smaller threshold) node MUST reset its
retransmission timers so, that if it employs exponential back-off,
the timers start over. After reaching short messages - usually below several hundred bytes. In
most cases the smallest allowed value for
fragmentation threshold implementation MUST continue probing using it
until either only time when IKE endpoints exchange completes or times out.

PMTU discovery messages of
several kBytes in IKE size is supposed to be coarse-grained, i.e. it is
expected, that node will try only few fragmentation thresholds, in
order to minimize possible IKE SA establishment delay. In a corner
case, when host will use only and often each
endpoint sends exactly one value, PMTU discovery will
effectively be disabled. In most cases such message.

For the reasons atriculated above implementing PMTU discovery will not be
needed, as in IKE
is OPTIONAL. It is believed that using values, the values recommended in section
Section 2.5.1, should
suffice. It 2.5.1 as fragmentation threshold will be sufficient in most
cases. Using these values could lead to suboptimal fragmentation,
but it is expected, that acceptable given the amount of traffic IKE produces.
Implementation may support PMTU discovery may if there are good reasons
to do it (for example if it is intended to be useful used in environments
where PMTU MTU size are smaller, than those is possible to be less that values listed in
Section 2.5.1, for example due to the presence of intermediate
tunnels. 2.5.1).

PMTU discovery in IKE follows recommendations, recommendations given in Section 10.4
of RFC4821 [RFC4821] with some differences, the difference, induced by the specialties of IKE. In particular:

o Unlike classical PMTUD [RFC1191] and PLMTUD [RFC4821] the goal of
Path MTU discovery in IKE
listed above. The difference is not to find the largest size of IP
packet, that will not be fragmented en route, but to find any
reasonable size, probably far from optimal.

o There the PMTU search is no goal to completely disallow IP fragmentation until its
presence leads to inability IKE to communicate (e.g. when IP
fragments are dropped)

o performed
downward, while in [RFC4821] it is performed upward. The reason for
this change is that IKE usually sends large messages only in IKE_AUTH exchange, i.e.
once per when IKE SA. Most of other messages will have size below
several hundred bytes. Performing full PMTUD for sending exactly
one large message SA
is inefficient.

In case of PMTU discovery Total Fragments field being established and in many cases there is used to
distinguish between different sets of fragments, i.e. only one such
message. If the sets that probing were obtained by fragmenting original performed upward this message would be
fragmented using different
fragmentation thresholds. As sender will start from larger fragments
and then make them smaller, the smallest allowable threshold, and usually all
other messages are small enough to avoid IP fragmentation, so there
would be little value to continue probing.

It is the Initiator of the exchange, who performs PMTU discovery. It
is done by probing several values of fragmentation threshold.
Implementation MUST be prepared to probe in every exchange that
utilizes IKE Fragmentation to deal with possible changes of path MTU
over time. While doing probes, it MUST start from larger values and
refragment original message using next smaller value of threshold if
it did not receive response in a reasonable time after several
retransmissions. The exact number of retransmissions and length of
timeouts are not covered in this specification because they do not
affect interoperability. However, the timeout interval is supposed
to be relatively short, so that unsuccessful probes would not delay
IKE operations too much. Performimg few retries within several
seconds for each probe seems appropriate, but different environments
may require different rules. When starting new probe node MUST reset
its retransmission timers so, that if it employs exponential back-
off, the timers will start over. After reaching the smallest allowed
value for fragmentation threshold implementation MUST continue
retransmitting until either exchange completes or times out using
timeout interval from Section 2.4 of [IKEv2].

PMTU discovery in IKE is supposed to be coarse-grained, i.e. it is
expected, that node will try only few fragmentation thresholds, in
order to minimize delays caused by unsuccessful probes. If no
information about path MTU is known yet, endpoint may start probing
from link MTU size. In the following exchanges node should start
from the current value of fragmentation threshold.

If implementation is capable to receive ICMP error messages it may
additionally utilize classic PMTU discovery methods, described in
[RFC1191] and [RFC1981]. In particular, if the Initiator receives
Packet Too Big error in response to the probe, and it contains
smaller value, than current fragmentation threshold, then the
Initiator SHOULD stop retransmitting the probe and SHOULD select new
value for fragmentation threshold that is less than or equal to the
value from the ICMP message and meets the requirements listed below.

In case of PMTU discovery Total Fragments field will
increase is used to
distinguish between different sets of fragments, i.e. the sets that
were created by fragmenting original message using different
fragmentation thresholds. Since sender starts from larger fragments
and then make them smaller, the value in Total Fragments field
increases with each new try. probe. When selecting next smaller value of for
fragmentation threshold, sender MUST ensure that the value in Total
Fragments field is really increased. This requirement should not
become be
a problem for the sender, as because PMTU discovery in IKE is supposed
to be coarse-grained, so difference between previous and next
fragmentation thresholds will should be significant anyway. The necessity
to distinguish between the sets is vital for receiver as since receiving
any
valid fragment from newer set will mean means that it have to start
reassembling over and not to mix fragments from different sets.

2.5.3. Fragmenting Messages containing unencrypted unprotected Payloads

Currently there are no one of IKEv2 Exchanges defines exchanges that define messages,
containing both
unencrypted unprotected payloads and payloads, protected by
Encrypted Payload.
But However IKEv2 doesn't forbid does not prohibit such messages.
construction. If some future IKEv2 extension defines such a message
and it needs to be fragmented, all unprotected payloads MUST be
placed in the first fragment, fragment (with Fragment Number field equal to 1),
along with Encrypted Fragment Payload, which MUST be present in any every
IKE Fragment
Message. Message and be the last payload in it.

Below is an example of fragmenting message, containing both encrypted protected
and unencrypted unprotected Payloads.

HDR(MID=n), PLD0, SK(NextPld=PLD1) {PLD1 ... PLDN}

Original Message

HDR(MID=n), PLD0, SKF(NextPld=PLD1, Frag#=1, TotalFrags=m) {...},
HDR(MID=n), SKF(NextPld=0, Frag#=2, TotalFrags=m) {...},
...
HDR(MID=n), SKF(NextPld=0, Frag#=m, TotalFrags=m) {...}

IKE Fragment Messages

Note,

Note that the size of each IP Datagram bearing IKE Fragment Messages
SHOULD NOT
should not exceed fragmentation threshold, including the very first,
which first one,
that contains unprotected Payloads. This will reduce the size of
Encrypted Fragment Payload content in the first IKE Fragment Message
to accommodate all unprotected Payloads. In extreme cases case Encrypted
Fragment Payload will contain no data, but it is still MUST must be present
in the message, because only its presence allows receiver to
distinguish IKE Fragment Message from regular
determine that sender have used IKE message. Fragmentation.

2.6. Receiving IKE Fragment Message

Receiver identifies IKE Fragment Message by the presence of Encrypted
Fragment Payload in it. Note, that In most cases it is possible for this payload
to be not the first (and the only) payload in the message (see
Section 2.5.3). But for all currently defined IKEv2 exchanges this
payload will be the first and the
only payload in the message. message, however this may not be true for some
hypothetical IKE exchanges (see Section 2.5.3)

Upon receiving IKE Fragment Message the following actions are
performed:

o Check message validity - in particular, check whether values of
Fragment Number and Total Fragments in Encrypted Fragment Payload
are valid. The following tests need to be performed.

* check that Fragment Number and Total Fragments fields are non-
zero

* check that Fragment Number field is less than or equal to Total
Fragments field

* if reassembling has already started, check that Total Fragments
field is equal to or greater than Total Fragments field in
fragments,
fragments that have already received been stored in the reassembling
queue

If any of this tests fails message MUST be silently discarded.

o Check, that this IKE Fragment Message is new for the receiver and
not a replay. If IKE Fragment message with the same Message ID,
same Fragment Number and same Total Fragments fields was is already
received and successfully processed,
present in the reassembling queue, this message is considered a
replay and MUST be silently discarded.

o Verify IKE Fragment Message authenticity by checking ICV in
Encrypted Fragment Payload. If ICV check fails message MUST be
silently discarded.

o If reassembling isn't is not finished yet and Total Fragments field in
received IKE Fragment Message fragment is greater than this field in
previously received those fragments,
that are in the reassembling queue, receiver MUST discard all
received fragments and start reassembling over with just received
IKE Fragment Message.

o Store message in the list reassembling queue waiting for the rest of
fragments to arrive.

When all IKE Fragment Messages (as indicated in the Total Fragments
field) are received, decrypted content of their already decrypted all Encrypted Fragment
Payloads is merged together to form content of original Encrypted
Payload, and, therefore, along with IKE Header and
unencrypted unprotected
Payloads (if any), original message. Then it is processed as if it
was received, verified and decrypted as regular
unfragmented IKE message.

If receiver doesn't does not get all IKE Fragment Messages fragments needed to reassemble
original Message for some Exchange within a timeout
interval, it acts according with Section 2.1 of [RFC5996], i.e.

retransmits interval, it MUST discard all
received so far IKE Fragment Messages for the exchange. Next actions
depend on the role of receiver in the exchange.

o The Initiator acts as described in Section 2.1 of [IKEv2]. It
either retransmits the fragmented request Message or deems IKE SA
to have failed and deletes it. The number of retransmits and
length of timeouts for the Initiator are not covered in this
specification since they are assumed to be the same as in regular
IKEv2 exchange and are discussed in Section 2.4 of [IKEv2].

o The Responder in this case acts as if no request message was
received. The reassembling timeout for Responder is RECOMMENDED
to be equal to the fragmented request Message (in case time interval that implementation waits before
completely giving up when acting as Initiator of Initiator) or
deems Exchange to have failed. If Exchange is abandoned, all
received so far IKE Fragment Messages exchange.
Section 2.4 of [IKEv2] gives recommendations for that Exchange MUST be
discarded. selecting this
interval. Implementation MAY use shorter timeout to conserve
memory.

2.6.1. Changes in Replay Protection Logic Detection and Retransmissions

According to [RFC5996] IKEv2 MUST [IKEv2] implementation must reject message with the same
Message ID as it has seen before (taking into consideration Response
bit). This logic has already been updated by [RFC6311], which
deliberately allows any number of messages with zero Message ID.
This document also updates this logic: if message contains Encrypted
Fragment Payload, the values of Fragment Number and Total Fragments
fields from this payload MUST be used along with Message ID to detect
retransmissions and replays.

If Responder receives IKE Fragment Message after it received,
successfully verified and processed regular message with the same
Message ID, it means that response message didn't reach Initiator and
it activated IKE Fragmentation. If Fragment Number in Encrypted
Fragment Payload in this message is equal to 1, Responder MUST
fragment its response and retransmit it back to Initiator in
fragmented form.

If Responder receives a replay IKE Fragment Message for already
reassembled, verified and processed fragmented message, it MUST
retransmit response back to Initiator, but only if Fragment Number
field in Encrypted Fragment Payload is equal to 1 and MUST silently
discard received message otherwise. If Total Fragments field in
received IKE Fragment Message is greater than in IKE Fragment
Messages that already processed fragmented message was reassembled
from, Responder MAY refragment its response message using smaller
fragmentation threshold before resending it back to Initiator. In has already been updated by [RFC6311], which
deliberately allows any number of messages with zero Message ID.
This document also updates this case Total Fragments field in new logic for the situations, when IKE Fragment Messages MUST be
greater than
Fragmentation is in previously sent IKE Fragment Messages. use.

If Initiator doesn't receive any incomimg message contains Encrypted Fragment Payload, the values
of response IKE Fragment Messages
within a timeout interval, it MAY refragment request Message using
smaller fragmentation threshold before retransmitting it (see
Section 2.5.1). In this case Number and Total Fragments field in new IKE
Fragment Messages fields MUST be greater than in previously sent IKE
Fragment Messages. Alternatively, if Initiator does receive some
(but not all) used along with
Message ID to detect retransmissions and replays.

If Responder receives retransmitted fragment of response IKE Fragment Messages, request when it MAY retransmit has
already processed that request and has sent back a response, that
event MUST only the first trigger retransmission of request IKE Fragment Messages, where the response message
(fragmented or not) if Fragment Number field in received fragment is equal
set to 1. 1 and MUST be ignored otherwise.

3. Interaction with other IKE extensions

IKE Fragmentation is compatible with most of defined IKE extensions,
like such as
IKE Session Resumption [RFC5723], ([RFC5723]), Quick Crash Detection Method
[RFC6290]
([RFC6290]) and so on. It neither affect their operation, nor is
affected by them. It is believed that IKE Fragmentation will also be
compatible with most future IKE extensions, if they follow general
principles of formatting, sending and receiving IKE messages,
described in [RFC5996]. [IKEv2].

When IKE Fragmentation is used with IKE Session Resumption [RFC5723],
([RFC5723]), messages of IKE_SESSION_RESUME Exchange cannot be
fragmented as since they
don't do not contain Encrypted Payload. These
messages may be large due to the ticket size. If this is the case the described solution won't help. To avoid IP
Fragmentation in this situation it is recommended to use smaller
tickets, e.g. by utilizing "ticket by reference" approach instead of
"ticket by value".

One exception that requires a special care is [RFC6311] - Protocol Support for
High Availability of IKEv2. As IKEv2/IPsec ([RFC6311]). Since it deliberately
allows any number of synchronization Exchanges exchanges to have the same
Message ID - ID, namely zero, standard IKEv2 replay detection logic, based
on checking Message ID is not applicable for such messages, and
receiver has to check message content to detect replays. When
implementing IKE Fragmentation along with [RFC6311], IKE Message ID
Synchronization messages MUST NOT be sent fragmented to simplify
receiver's task of detecting replays. Fortunately, these messages
are small and there is no point in fragmenting them anyway.

4. Transport Considerations

With IKE Fragmentation if any single IKE Fragment Message get lost,
receiver becomes unable to reassemble original Message. So, in
general, using IKE Fragmentation implies higher probability for the
Message not to be delivered to the peer. Although in most network
environments the difference will be insignificant, on some lossy
networks it may become noticeable. When using IKE Fragmentation
implementations MAY use longer timeouts and do more retransmits than
usual before considering peer dead.

Note that Fragment Messages are not individually acknowledged. The
response Fragment Messages are sent back all together only when all
fragments of request are received, the original request Message is
reassembled and successfully processed.

5. Security Considerations

Most of the security considerations for IKE Fragmentation are the
same as those for the base IKEv2 protocol described in [RFC5996]. [IKEv2]. This
extension introduces Encrypted Fragment Payload to protect content of
IKE Message Fragment. This allows receiver to individually check
authenticity of fragments, thus protecting peers from DoS attack.

Security Considerations Section of [RFC5996] [IKEv2] mentions possible attack
on IKE by exhausting of the IP reassembly buffers. The mechanism,
described in this document, allows IKE to avoid IP-fragmentation IP fragmentation and
therefore increases its robustness to DoS attacks.

The following attack is possible with IKE Fragmentation. An attacker
can initiate IKE_SA_INIT exchange, Exchange, complete it, compute SK_a and SK_e
and then send a large, but still incomplete, set of IKE_AUTH
fragments. These fragments will pass the ICV check and will be
stored in reassembly buffers, but as since the set is incomplete, the
reassembling will never succeed and eventually will time out. If the
set is large, this attack could potentially exhaust the receiver's
memory resources.

To mitigate the impact of this attack, it is RECOMMENDED that
receiver limits the number of fragments it stores in reassembling
queue so that the sum of the sizes of Encrypted Fragment Payload
contents (after decryption) for fragments that are already placed
into the reassembling queue be less than some value that is
reasonable for the implementation. If the peer sends so many
fragments, that the above condition is not met, the receiver can
consider this situation to be either attack or as broken sender
implementation. In either case, the receiver SHOULD drop the
connection and discard all the received fragments.

This value can be predefined, can be a configurable option, or can be
calculated dynamically depending on receiver's memory load. In any
case, the value SHOULD NOT exceed 64 Kbytes (the maximum size of UDP
datagram) because any IKE message before fragmentation must be
shorter than that.

6. IANA Considerations

This document defines new Payload in the "IKEv2 Payload Types"
registry:

<TBA> Encrypted and Authenticated Fragment SKF

This document also defines new Notify Message Types in the "Notify
Message Types - Status Types" registry:

<TBA> IKEV2_FRAGMENTATION_SUPPORTED

7. Acknowledgements

The author would like to thank Tero Kivinen, Yoav Nir, Paul Wouters,
Yaron Sheffer, Joe Touch, Derek Atkins Atkins, Ole Troan and others for
their reviews and valuable comments. Thanks to Ron Bonica for
contributing text to the Introduction Section. Thanks to Paul
Hoffman and Barry Leiba for improving text clarity.

8. References

8.1. Normative References

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

[RFC5996]

[IKEv2] Kaufman, C., Hoffman, P., Nir, Y., and P. Eronen, P., and T.
Kivinen, "Internet Key Exchange Protocol Version 2
(IKEv2)",
RFC 5996, September 2010. draft-kivinen-ipsecme-ikev2-rfc5996bis-03 (work
in progress), April 2014.

[RFC6311] Singh, R., Kalyani, G., Nir, Y., Sheffer, Y., and D.
Zhang, "Protocol Support for High Availability of IKEv2/
IPsec", RFC 6311, July 2011.

8.2. Informative References

[RFC0791] Postel, J., "Internet Protocol", STD 5, RFC 791,
September 1981.

[RFC1191] Mogul, J. and S. Deering, "Path MTU discovery", RFC 1191,
November 1990.

[RFC1981] McCann, J., Deering, S., and J. Mogul, "Path MTU Discovery
for IP version 6", RFC 1981, August 1996.

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

[RFC4787] Audet, F. and C. Jennings, "Network Address Translation
(NAT) Behavioral Requirements for Unicast UDP", BCP 127,
RFC 4787, January 2007.

[RFC4821] Mathis, M. and J. Heffner, "Packetization Layer Path MTU
Discovery", RFC 4821, March 2007.

[RFC5282] Black, D. and D. McGrew, "Using Authenticated Encryption
Algorithms with the Encrypted Payload of the Internet Key
Exchange version 2 (IKEv2) Protocol", RFC 5282,
August 2008.

[RFC5405] Eggert, L. and G. Fairhurst, "Unicast UDP Usage Guidelines
for Application Designers", BCP 145, RFC 5405,
November 2008.

[RFC5723] Sheffer, Y. and H. Tschofenig, "Internet Key Exchange
Protocol Version 2 (IKEv2) Session Resumption", RFC 5723,
January 2010.

[RFC6290] Nir, Y., Wierbowski, D., Detienne, F., and P. Sethi, "A
Quick Crash Detection Method for the Internet Key Exchange
Protocol (IKE)", RFC 6290, June 2011.

[RFC6888] Perreault, S., Yamagata, I., Miyakawa, S., Nakagawa, A.,
and H. Ashida, "Common Requirements for Carrier-Grade NATs
(CGNs)", BCP 127, RFC 6888, April 2013.

[FRAGDROP]
Jaeggli, J., Colitti, L., Kumari, W., Vyncke, E., Kaeo,
M., and T. Taylor, "Why Operators Filter Fragments and
What It Implies", draft-taylor-v6ops-fragdrop-02 (work in
progress), December 2013.

[BLACKHOLES]
De Boer, M. and J. Bosma, "Discovering Path MTU black
holes on the Internet using RIPE Atlas", July 2012, <http:
//www.nlnetlabs.nl/downloads/publications/
pmtu-black-holes-msc-thesis.pdf>.

[DOSUDPPROT]
Kaufman, C., Perlman, R., and B. Sommerfeld, "DoS
protection for UDP-based protocols", ACM Conference on
Computer and Communications Security, October 2003.

Appendix A. Design rationale

The simplest approach to the IKE fragmentation would have been to
fragment message that is fully formed and ready to be sent. But if
message got fragmented after being encrypted and authenticated, this
could open a possibility for a simple Denial of Service attack. The
attacker could infrequently emit forged but valid looking fragments
into the network, and some of these fragments would be fetched by
receiver into the reassembling queue. Receiver could not distinguish
forged fragments from valid ones and could only determine that some
of received fragments were forged when the whole message got
reassembled and check for its authenticity failed.

To prevent this kind of attack and also to reduce vulnerability to
some other kinds of DoS attacks it was decided to make fragmentation
before applying cryptographic protection to the message. In this
case each Fragment Message becomes individually encrypted and
authenticated, that allows receiver to determine forged fragments and
not to store them in the reassembling queue.

Appendix B. Correlation between IP Datagram size and Encrypted Payload
content size

For IPv4 Encrypted Payload content size is less than IP Datagram size
by the sum of the following values:

o IPv4 header size (typically 20 bytes, up to 60 if IP options are
present)

o UDP header size (8 bytes)

o non-ESP marker size (4 bytes if present)

o IKE Header size (28 bytes)

o Encrypted Payload header size (4 bytes)

o IV size (varying)

o padding and its size (at least 1 byte)

o ICV size (varying)

The sum may be estimated as 61..105 bytes + IV + ICV + padding.

For IPv6 Encrypted Payload content size is less than IP Datagram size
by the sum of the following values:

o IPv6 header size (40 bytes)

o IPv6 extension headers (optional, size varies)

o UDP header size (8 bytes)

o non-ESP marker size (4 bytes if present)

o IKE Header size (28 bytes)

o Encrypted Payload header size (4 bytes)

o IV size (varying)

o padding and its size (at least 1 byte)

o ICV size (varying)

If no extension header is present, the sum may be estimated as 81..85
bytes + IV + ICV + padding. If extension headers are present, the
payload content size is further reduced by the sum of the size of the
extension headers. The length of each extension header can be
calculated as 8 * (Hdr Ext Len) bytes except for the fragment header
which is always 8 bytes in length.

Author's Address

Valery Smyslov
ELVIS-PLUS
PO Box 81
Moscow (Zelenograd) 124460
Russian Federation

Phone: +7 495 276 0211
Email: svan@elvis.ru