Internet DRAFT - draft-mavrogiannopoulos-tls-cid
draft-mavrogiannopoulos-tls-cid
TLS Working Group N. Mavrogiannopoulos
Internet-Draft RedHat
Intended status: Standards Track H. Tschofenig
Expires: November 17, 2017 ARM
T. Fossati
Nokia
May 16, 2017
Datagram Transport Transport Layer Security (DTLS) Transport-Agnostic
Security Association Extension
draft-mavrogiannopoulos-tls-cid-01
Abstract
This memo proposes a new Datagram Transport Transport Layer Security
(DTLS) extension for DTLS 1.2 that provides the ability to negotiate,
during handshake, a transport independent identifier that is unique
per security association. This identifier effectively decouples the
DTLS session from the underlying transport protocol, allowing the
same security association to be migrated across different instances
of the same transport, or to a completely different transport.
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 November 17, 2017.
Copyright Notice
Copyright (c) 2017 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
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 1]
Internet-Draft DTLS ta_sa Extension May 2017
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 . . . . . . . . . . . . . . . . . . . . . . . . 2
2. Conventions used in this document . . . . . . . . . . . . . . 3
3. Transport Agnostic Security Associatiation Extension . . . . 4
3.1. Extended Client Hello . . . . . . . . . . . . . . . . . . 4
3.2. Extended Server Hello . . . . . . . . . . . . . . . . . . 5
3.3. Handling unknown CIDs . . . . . . . . . . . . . . . . . . 6
3.4. Wire Format Changes . . . . . . . . . . . . . . . . . . . 6
3.5. De-duplication Algorithm . . . . . . . . . . . . . . . . 7
4. Clashing HOTP CIDs . . . . . . . . . . . . . . . . . . . . . 7
5. Security Considerations . . . . . . . . . . . . . . . . . . . 8
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 8
7. Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . 8
8. References . . . . . . . . . . . . . . . . . . . . . . . . . 8
8.1. Normative References . . . . . . . . . . . . . . . . . . 9
8.2. Informative References . . . . . . . . . . . . . . . . . 9
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 10
1. Introduction
DTLS security context demultiplexing is done via the 5-tuple.
Therefore, the security association needs to be re-negotiated from
scratch whenever the transport identifiers change. For example, when
moving the network attachment from WLAN to a cellular connection, or
when the IP address of the IoT devices changes during a sleep cycle.
A NAT device may also modify the source UDP port after a short idle
period. In such cases, there is not enough information in the DTLS
record header for a server that is handling multiple concurrent
sessions to associate the new address to an existing client.
This memo proposes a new TLS extension [RFC6066] for DTLS 1.2 that
provides the ability to negotiate, at handshake time, a transport
independent identifier that is unique per security association. We
call this identifier Connection ID (CID). Its function is to
effectively decouple the DTLS session from the underlying transport
protocol, allowing the same DTLS security association to be migrated
across different instances of the same transport, or even to a
completely different transport - e.g., from UDP to GSM-SMS as showed
in Figure 1.
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 2]
Internet-Draft DTLS ta_sa Extension May 2017
00
/\
:
IP UDP : DTLS Record Header
+-----+-----+-------+ +-----+-----+ : +---------+-------+------
| src | dst | proto | | src | dst | : | Seq#i | CID | ...
+-----+-----+-------+ +-----+-----+ : +---------+-------+------
`----------------+----------------' : ^
`................ : .............'
<Handover event> :
GSM-SMS : DTLS Record Header
+-------+-------+ : +---------+-------+-----
| tp-oa | tp-da | : | Seq#i+1 | CID | ...
+-------+-------+ : +---------+-------+-----
:
\/
00
Figure 1: Transparent Handover of DTLS Session
We present two methods for producing the CID: the first uses a single
value generated unilaterally by the server which is fixed throughout
the session, whereas the second provides a sequence of identifiers
that are created using a HMAC-based OTP algorithm [RFC4226] keyed
with a per-session shared secret (see Section 3.1 for details). The
latter allows a client to shift to a new identifier, for example when
switching networks, and is intended as a mechanism to counteract
tracking by third party observers. However, it must be noted that
this is not generally applicable as a tracking-protection measure: in
fact, it becomes totally ineffective when the client is oblivious of
changes in the underlying transport identifiers (e.g., on NAT rebind
after timeout), and also does not guarantee unique identifiers (see
Section 4 for further details). Both methods generate a CID that is
32-bits in size, like the Security Parameter Index (SPI) in IPsec
[RFC4301].
Similar approaches to support transparent handover of a DTLS session
have been described in [I-D.barrett-mobile-dtls] and [DTLSMOB].
2. Conventions used in this document
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in
[RFC2119].
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 3]
Internet-Draft DTLS ta_sa Extension May 2017
3. Transport Agnostic Security Associatiation Extension
In order to negotiate a Transport Agnostic Security Association,
clients include an extension of type "ta_sa" in the extended client
hello (Section 3.1). Servers that receive an extended hello
containing a "ta_sa" extension MAY agree to use a Transport Agnostic
Security Association by including an extension of type "ta_sa" in the
extended server hello (Section 3.2).
If both server and client agree, the DTLSCiphertext format does
change after the DTLS connection state is updated; i.e.: for the
sending side, after the ChangeCipherSpec message is sent, for the
receiving sides, after the ChangeCipherSpec is received.
The DTLSCiphertext format is changed for both the client and the
server. However, only a client can initiate a switch to an unused
'cid' value; a server MUST utilize the same value seen on the last
valid message received by the client.
struct {
ContentType type;
ProtocolVersion version;
uint16 epoch;
uint48 sequence_number;
uint32 cid; // New field
uint16 length;
select (CipherSpec.cipher_type) {
case block: GenericBlockCipher;
case aead: GenericAEADCipher;
} fragment;
} DTLSCiphertext;
Figure 2: Modified DTLS Record Format
See Section 3.3 for details on how to handle receipt of unknown or
unexpected CIDs.
3.1. Extended Client Hello
In order to negotiate a Transport Agnostic Security Association,
clients include an extension of type "ta_sa" in the extended client
hello. The "extension_data" field of this extension SHALL contain
the ClientSecAssocData structure in Figure 3.
In case the fixed(0) type has been negotiated, the 'cid' of the
packets after ChangeCipherSpec is sent explicitly by the server.
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 4]
Internet-Draft DTLS ta_sa Extension May 2017
In case the hotp(1) type has been negotiated, the initial 'cid' is
calculated using the HOTP algorithm ([RFC4226]) as follows:
o A 20-byte string is generated using a [RFC5705] exporter. The key
material exporter uses the label "EXPORTER-ta-security-
association-hotp" without the quotes, and without any context
value.
o The initial 'cid' equals to the first HOTP value (i.e., the 31-bit
value of Sbits in [RFC4226] notation), generated by using the
previously exported value as K.
Subsequent values of the HOTP algorithm can be used in place of the
initial, as long as they fall into the negotiated window_size (see
Figure 4).
enum {
fixed(0), hotp(1), (255)
} SecAssocType;
struct {
SecAssocType types<1..2^8-1>;
} ClientSecAssocData;
Figure 3: ta_sa extension, client
3.2. Extended Server Hello
Servers that receive an extended hello containing a "ta_sa" extension
MAY agree to use a Transport Agnostic Security Association by
including an extension of type "ta_sa", with "extension_data" being
ServerSecAssocData in the extended server hello (Figure 4).
struct {
SecAssocType type;
select (type) {
case fixed:
struct {
uint32 cid_value;
};
case hotp:
struct {
uint16 window_size;
};
};
} ServerSecAssocData;
Figure 4: ta_sa extension, server
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 5]
Internet-Draft DTLS ta_sa Extension May 2017
In case the fixed(0) type is chosen, 'cid_value' contains the value
to be used as 'cid'. In case hotp(1) type is chosen, 'window_size'
must be greater or equal to 1, indicating the number of HOTP values
that the server can recognize for this particular client.
3.3. Handling unknown CIDs
Server might need to deal with unknown or unexpected CIDs.
An unknown CID is one that is not found in the server's lookup table.
This could happen because:
o Server reboots and looses its state; or
o There is a genuine bug in either client or server code (or even
somewhere on the network path).
Either way, the server will not be able to locate a suitable context
to use for sending an (encrypted) Alert back to the sender.
Therefore, the server should simply discard records containing an
unknown CID.
A CID might be unexpected if it existed but it's been already
shifted. This situation may arise if the packet has been delayed by
the network. Server MAY ignore a record carrying an unexpected CID.
Ignoring unknown or unexpected CIDs should also reduce the attack
surface.
3.4. Wire Format Changes
How to signal the modified wire format to the receiving end is
currently an open problem.
Note that moving the cid after the length field and computing the
difference between the UDP datagram's and DTLS record's lengths is
not an option because there is no guarantee that UDP datagrams carry
one and one only DTLS record (Section 4.1.1. of [RFC6347]).
Ideally, we would just bump the version number, but there seems to be
limited room for maneuver given the way TLS encodes version
information in the record header, and also given that we want CID to
work with DTLS 1.2 and later.
More discussion needed to sort out this point.
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 6]
Internet-Draft DTLS ta_sa Extension May 2017
3.5. De-duplication Algorithm
The following algorithm assumes that receivers have an unambiguous
way to tell that the wire format is the one described in Section 3.
As suggested in Section 3.4, this could be signalled by a different
version number { TBD, TBD }.
In order to enqueue an incoming record to the right security context,
a receiver SHALL extract the 4-tuple from the UDP packet and the
ContentType of the TLS record. Then, if ContentType is
change_cipher_spec or handshake, the receiver SHALL use the sender
address to lookup or create (if ContentType is handshake and
HandshakeType is one of ClientHello or ServerHello) the session
context. If ContentType is one of application_data or alert,
receiver SHALL inspect the ProtocolVersion field and: if it is { 3, 3
} (i.e., TLS v1.2), use the 4-tuple to lookup the session context, if
it is { TBD, TBD }, extract the CID from the record and use it to
lookup the session context.
4. Clashing HOTP CIDs
HOTP behaves like a PRF, thus uniformly distributing the produced
CIDs across the 32-bit space. Table 1 presents the probability to
end up with two separate sessions having the same HOTP CID when the
number of concurrent sessions and/or the length of the CIDs sequence
is increased.
+-----------------------+-------------------------------------------+
| Sessions x | Collision probability |
| window_size | |
+-----------------------+-------------------------------------------+
| 10 | 1.16415320717e-08, or about 1 in |
| | 85,899,347 |
| 100 | 1.16415254059e-06, or about 1 in 858,994 |
| 1000 | 0.000116408545826, or about 1 in 8,590 |
| 10000 | 0.011574031737, or about 1 in 86 |
| 100000 | 0.687813095694, or about 1 in 1 |
| 1000000 | 1.0, or about 1 in 1 |
+-----------------------+-------------------------------------------+
Table 1
The takeaway is that 32-bits are probably too few for highly loaded
servers that want to do HOTP as their primary CID allocation
strategy. An alternative would be for the server to stop negotiating
'hotp' and fall back to 'fixed' when the number of active sessions
crosses some threshold; another would be to increase the CID space to
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 7]
Internet-Draft DTLS ta_sa Extension May 2017
40 or 48 bits when HOTP is used; yet another would be to allow the
length to be negotiable.
5. Security Considerations
CID does not affect the running protocol in any way other than adding
an un-authenticated field to the record header. As such, this
identifier has no effect on the overall security of the session with
respect to authentication, confidentiality and integrity. On the
other hand, since this identifier is not authenticated, it should not
be used in any way that assumes it is, nor be assumed to be secret or
unknown to an adversary. In general, this identifier should not be
relied on more than the IP address or UDP port numbers are.
To address the privacy concerns of using a fixed identifier for the
lifetime of a session which may roam through multiple networks, we
have introduced the hotp identifier type. This type of identifier
gives the client a chance to switch its ts_sa identity when also
switching its transport identifiers or network attachment (assuming
that client is made aware of the change before it sends a new DTLS
record). The choice of which type of identifier to use is a trade-
off between the request for privacy stated by the client and the
ability of the server to control the identifiers in use at each point
in time, as explained in Section 4.
6. IANA Considerations
This document adds a new extension for DTLS: ts_sa(TODO). This
extension MUST only be used with DTLS, and not with TLS. This
extension is assigned from the TLS ExtensionType registry defined in
[RFC5246].
7. Acknowledgments
Thanks to Achim Kraus, Carsten Bormann, Kai Hudalla, Simon Bernard,
Stephen Farrell, for helpful comments and discussions that have
shaped the document.
This work is partially supported by the European Commission under
Horizon 2020 grant agreement no. 688421 Measurement and Architecture
for a Middleboxed Internet (MAMI). This support does not imply
endorsement.
8. References
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 8]
Internet-Draft DTLS ta_sa Extension May 2017
8.1. Normative References
[I-D.ietf-tls-tls13]
Rescorla, E., "The Transport Layer Security (TLS) Protocol
Version 1.3", draft-ietf-tls-tls13-20 (work in progress),
April 2017.
[I-D.rescorla-tls-dtls13]
Rescorla, E., Tschofenig, H., and N. Modadugu, "The
Datagram Transport Layer Security (DTLS) Protocol Version
1.3", draft-rescorla-tls-dtls13-01 (work in progress),
March 2017.
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<http://www.rfc-editor.org/info/rfc2119>.
[RFC4226] M'Raihi, D., Bellare, M., Hoornaert, F., Naccache, D., and
O. Ranen, "HOTP: An HMAC-Based One-Time Password
Algorithm", RFC 4226, DOI 10.17487/RFC4226, December 2005,
<http://www.rfc-editor.org/info/rfc4226>.
[RFC5246] Dierks, T. and E. Rescorla, "The Transport Layer Security
(TLS) Protocol Version 1.2", RFC 5246,
DOI 10.17487/RFC5246, August 2008,
<http://www.rfc-editor.org/info/rfc5246>.
[RFC5705] Rescorla, E., "Keying Material Exporters for Transport
Layer Security (TLS)", RFC 5705, DOI 10.17487/RFC5705,
March 2010, <http://www.rfc-editor.org/info/rfc5705>.
[RFC6066] Eastlake 3rd, D., "Transport Layer Security (TLS)
Extensions: Extension Definitions", RFC 6066,
DOI 10.17487/RFC6066, January 2011,
<http://www.rfc-editor.org/info/rfc6066>.
[RFC6347] Rescorla, E. and N. Modadugu, "Datagram Transport Layer
Security Version 1.2", RFC 6347, DOI 10.17487/RFC6347,
January 2012, <http://www.rfc-editor.org/info/rfc6347>.
8.2. Informative References
[DTLSMOB] Seggelmann, R., Tuexen, M., and E. Rathgeb, "DTLS
Mobility", 2012.
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 9]
Internet-Draft DTLS ta_sa Extension May 2017
[I-D.barrett-mobile-dtls]
Williams, M. and J. Barrett, "Mobile DTLS", draft-barrett-
mobile-dtls-00 (work in progress), March 2009.
[RFC4301] Kent, S. and K. Seo, "Security Architecture for the
Internet Protocol", RFC 4301, DOI 10.17487/RFC4301,
December 2005, <http://www.rfc-editor.org/info/rfc4301>.
Authors' Addresses
Nikos Mavrogiannopoulos
RedHat
EMail: nmav@redhat.com
Hannes Tschofenig
ARM
EMail: hannes.tschofenig@arm.com
Thomas Fossati
Nokia
EMail: thomas.fossati@nokia.com
Mavrogiannopoulos, et alExpires November 17, 2017 [Page 10]
