Internet Engineering Task Force A. Vives
Internet-Draft J. Palet
Expires: April 24, 2005 Consulintel
P. Savola
CSC/FUNET
October 24, 2004
IPv6 Security Problem Statement
draft-vives-v6ops-ipv6-security-ps-02.txt
Status of this Memo
This document is an Internet-Draft and is subject to all provisions
of section 3 of RFC 3667. By submitting this Internet-Draft, each
author represents that any applicable patent or other IPR claims of
which he or she is aware have been or will be disclosed, and any of
which he or she become aware will be disclosed, in accordance with
RFC 3668.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as
Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six months
and may be updated, replaced, or obsoleted by other documents at any
time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
This Internet-Draft will expire on April 24, 2005.
Copyright Notice
Copyright (C) The Internet Society (2004).
Abstract
Today, each network is often secured by a unique device (i.e.
security gateway or firewall) that becomes a bottleneck for the
end-to-end security model with IPv6. The deployment of IPv6 enabled
devices and networks bring some issues, which must be addressed by
security administrators in order to guarantee at least the same level
Vives, et al. Expires April 24, 2005 [Page 1]
Internet-Draft IPv6 Security Problem Statement October 2004
of security that is obtained nowadays with IPv4 and network-based
(including perimeter-based) security schemes, allowing at the same
time all the IPv6 advantages.
The most important issues are the rediscovery of end-to-end
communications, the availability of IPsec in all IPv6 stacks, the
increase in the number and type of IP devices and also the increase
in the number of "nomadic" devices, connecting to different networks
that could have different security policies.
The security policies and architectures currently applied in Internet
with IPv4 do no longer apply for end-to-end security models which
IPv6 will need. This document outlines the advantages and drawbacks
of both security schemes: network/perimeter-based and distributed.
This document aims to identify IPv6 issues that justify the need of a
distributed security model for IPv6, that is, simply to show that a
security problem will arise with the deployment of IPv6 networks if
nothing is done.
Vives, et al. Expires April 24, 2005 [Page 2]
Internet-Draft IPv6 Security Problem Statement October 2004
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Network-based versus Host-based Security . . . . . . . . . . . 5
2.1 Network-based Security . . . . . . . . . . . . . . . . . . 5
2.2 Host-based Security . . . . . . . . . . . . . . . . . . . 7
3. IPv6 Issues . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.1 End-to-End . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2 IPsec-encrypted ESP-traffic in transport mode . . . . . . 11
3.3 Mobility . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4 Addresses . . . . . . . . . . . . . . . . . . . . . . . . 11
3.4.1 Link-local addresses . . . . . . . . . . . . . . . . . 12
3.4.2 New Multicast addresses . . . . . . . . . . . . . . . 12
3.5 Neighbor Discovery Weakness . . . . . . . . . . . . . . . 13
3.6 Routing Header . . . . . . . . . . . . . . . . . . . . . . 13
3.7 Home Address Option . . . . . . . . . . . . . . . . . . . 13
3.8 Embedded Devices . . . . . . . . . . . . . . . . . . . . . 14
4. Other Issues . . . . . . . . . . . . . . . . . . . . . . . . . 14
5. Security Considerations . . . . . . . . . . . . . . . . . . . 14
6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 14
7. References . . . . . . . . . . . . . . . . . . . . . . . . . . 14
7.1 Normative References . . . . . . . . . . . . . . . . . . . . 14
7.2 Informative References . . . . . . . . . . . . . . . . . . . 14
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . 15
Intellectual Property and Copyright Statements . . . . . . . . 17
Vives, et al. Expires April 24, 2005 [Page 3]
Internet-Draft IPv6 Security Problem Statement October 2004
1. Introduction
This document will cope only with IPv6 issues related to security,
i.e., it will try to answer the following question: How would the
deployment of IPv6 affect the security of a network? This network
could be a dual-stack network with IPv6 traffic from IPv6 capable
nodes, or an IPv6 only network.
The deployment of IPv6 enabled devices and networks forces the
security administrator to consider several issues:
o The rediscovery of end-to-end communications.
o The availability of IPsec in all IPv6 stacks.
o The increase in the number and type of IP devices.
o The increase in the number of "nomadic" devices, connecting and
moving between different networks.
The security policies and architectures currently applied in Internet
with IPv4 no longer apply for end-to-end security models which IPv6
will enable. This document outlines the advantages and drawbacks of
both the security schemes: network/perimeter-based and distributed.
Also IPv6 issues will be identified that justify the need of
distributed security for IPv6, that is, simply to show that a
security problem will arise with the deployment of IPv6 networks if
traditional schemes are used.
The following issues are out of scope of this document and will be
addressed elsewhere:
o State the security requirements for the described IPv6 scenarios.
o Propose a solution or architecture to address the problem stated
in this document.
o To address security problems derived from the use of transition
mechanisms.
Last but not least, this document contains a brief definition of what
we understand by "security". We use security in the "big scope" of
the word, trying to include as much as possible. In other words, a
host, a network or some information, will be secure when no attacks
could succeed against them. A success will mean compromise of
availability, integrity, confidentiality or authenticity. The
realistic objective is to be as much secure as possible in a precise
Vives, et al. Expires April 24, 2005 [Page 4]
Internet-Draft IPv6 Security Problem Statement October 2004
moment. It will be part of the requirements to establish which kind
of security is given using a number of mechanisms.
For clarity, in the rest of this document, network-based security
includes also the perimeter-based model.
2.. Network-based versus Host-based Security
In this section two different approaches are analyzed to be used
later in the rationale about the security problems that IPv6 could
introduce
2.1 Network-based Security
This is the most used scheme, where the security of a host depends on
the point of the network it is connected to.
The perimeter scheme is the simplest one (see Fig. 1) and is based
in the topology of the network. The security policy is enforced in a
central host or firewall (FW), which provides secure network
connectivity to one or more network segments. The FW will be what an
"outside" host sees when tries to attack the network. Attacks coming
from the same LAN segment are not protected by the FW. Different
nodes (even different addresses in the same node) may have different
policies.
In a more advanced form of perimeter security, the different networks
could be protected from each other, or a number of internal firewalls
could be used as well. This way some networks could be protected
from hosts in other internal network
Vives, et al. Expires April 24, 2005 [Page 5]
Internet-Draft IPv6 Security Problem Statement October 2004
/-------\
/ \
| Internet |
\ /
\---+---/
|
| Policy Enforcement Point
+---+---+
LAN-1 | | DMZ-1
----+---------+ FW +-------+----
| | | |
+-+--+ +---+---+ +--+-+
| H1 | | | S1 |
+----+ |LAN-2 +----+
|
+----+ |
| H2 +--+
+----+ |
Figure 1: Perimeter Security
This model is based on the following assumptions:
o The threats come from "outside" the FW, basically the Internet.
o Everybody from the same LAN segment is trusted.
o The protected nodes won't go "outside" where FW won't be able to
protect them.
o There are no backdoors on the network (modem, WLAN, other
connections).
o The hosts will not need to be accessed directly from outside (at
least in a general manner, i.e., potentially all ports on all
hosts).
The main advantage of this scheme is its simplicity and easiness as
the elements and points of configuration are reduced to the minimum,
requiring few/no protocols and mechanisms to implement the security.
In case of a more complex configuration, where multiple FW are
deployed within an organization network, the complexity will
increase.
The drawbacks of this model are:
o This is a centralized model: Single point of failure for both
Vives, et al. Expires April 24, 2005 [Page 6]
Internet-Draft IPv6 Security Problem Statement October 2004
performance and availability. If the FW fails, then all the
networks connected to it loose network connectivity unless
specific fail-over techniques are applied.
o A big percentage of the threats come from inside the FW, and are
not addressed by this security model, especially when internal
firewalls are not deployed.
o The most dangerous threats, in the sense that one may not be able
to protect from them, come from inside the FW.
o The FW usually acts as NAT and/or proxy box, interfering or even
disallowing end-to-end communications. In complex configurations,
even more than one level of NAT/proxy could appear.
o Transport mode secured communications (using IPsec ESP for
example) need special solutions ([1]).
o The same security policy may be enforced for all the nodes of each
network connected to the FW, but it is also possible to have
separate policies for all hosts. In any case, an error in the FW
will equally expose all hosts in a network.
o Virtual organizations, for example those using GRID models, don't
work with traditional centralized security models.
o The lack of secure end-to-end prevents innovation.
2.2 Host-based Security
Host based security model, already introduced by [2], is based on the
idea of enforcing the security policy in each network host from a
central control point.
The three main elements identified in the distributed security model
are:
o Policy Specification Language.
o Policy Exchange Protocol.
o Authentication of Entities.
The basic idea is simple: the Security Policy is centrally defined
using the Policy Specification Language and distributed to each host
by means of the Policy Exchange Protocol. The Network Entities need
to be authenticated in order to be trusted, for example to allow an
Vives, et al. Expires April 24, 2005 [Page 7]
Internet-Draft IPv6 Security Problem Statement October 2004
incoming connection or to trust the received Security Policy.
The biggest challenge, however, is trusting that the hosts comply to
the rules they've received, for example that the user can't just
disable the firewall if (s)he dislike the policy; of course, this
only can happen in the case (s)he has administrative rights for that
(often not the case in non-personal systems, those not owned by the
end user). It seems that one ore more network entities would have to
keep watch over the hosts in order to detect if they are not
following the received policy. At first look the more appropriated
entity seems to be one that knows the security policy, for example
the one that distributes it.
From a security point of view this model somehow eases the work to
the "enemies", putting the Policy Enforcement Point in their hands.
So not only mechanisms to prevent direct attacks to the security
solution must be developed but mechanisms to minimize the
consequences.
/-------\
/ \
| Internet |
+------+ \ /
|Sec. | \---+---/
|Policy| |
+--+---+ |
| /-+-\
| | \ / | LAN-3
-+---+------+ x +-------+--
LAN-1 | (*) | / \ | | (*)Policy Enforcement Point
+--+-+ \-+-/ +-+--+
| H1 | | LAN-2 | H3 |
+----+ | +----+
| (*)
+--+-+
| H2 |
+----+
Figure 2: Host-based Security
This model is based on the following assumptions:
o Each host can be uniquely and securely identified.
o The security policy could be applied in one or more of the
following levels: network, transport and application.
Vives, et al. Expires April 24, 2005 [Page 8]
Internet-Draft IPv6 Security Problem Statement October 2004
o The threat comes from anywhere in the network.
o The intruder has no physical access to the protected network hosts
(what about malicious users? See other topics section).
o "Outside" hosts may be able to access all hosts "Inside",
depending on the policies.
The advantages of this model are:
o The security policy can be host-specific.
o A host can take better decisions as it knows what it is doing or
trying to do, that means it can better detect strange packets.
For example, it could allow mail traffic to only one application
on the system.
o Enables the usage of end-to-end applications level security (e.g.,
web services security standards).
o Enables better protection from attacks by the "internal" users,
and possibly even to a degree from those in the local segment.
For example address spoofing can be detected and avoided coming
from the same LAN segment, without router participation.
o Can protect a host independent of the topology, i.e., wherever the
host is connected.
o Does not need specific devices to secure a host (consider the case
of a single host with a CPE, if the CPE had no (user-controllable)
firewall functions).
o Can control the outgoing attempts from each host, avoiding local
network misbehavior or malicious practices.
o The collection of audit information could be more complete in a
distributed model, despite the processing of that information is
done in a distributed or centralized fashion.
o It maintains the centralized control of the security policies,
from where they are distributed to each host (central decisions,
local enforcement).
The drawbacks of this model are:
o It is more complex than the perimeter one.
o The uniqueness and secured identification of hosts is not trivial,
Vives, et al. Expires April 24, 2005 [Page 9]
Internet-Draft IPv6 Security Problem Statement October 2004
for example using certificates [2].
o The hosts must be trusted (or designed appropriately) so that they
will operate according to the policy. For example, it must be
impossible to disable the firewall functions or if the policy is
not followed network communication is not allowed.
o A host that becomes compromised or infected with a worm or virus
in any case can't be trusted to operate according to the policies,
as the worms/viruses probably first create holes or disable the
protections if they can.
o It may be challenging to design the system so that policy updates
are made available to the nodes which may not be network-reachable
all the time.
o It may be difficult to distinguish a misbehaving application from
a legitimate application (for example, many email worms may be
channeled through the MUA which must be authorized to send the
mails to operate correctly).
3. IPv6 Issues
When IPv6 is deployed, either in an existing IPv4 network or in a new
IPv6-only network, the security administrator must take into account
that IPv6 traffic will be different from the IPv4.
IPv6 enabled nodes will likely have global addresses, which means
they may be reachable from any other IPv6 node in the Internet. A
security administrator can prevent this by using local addressing
and/or firewalls, but the benefits of IPv6 may not be fully realized
if so.
The differences between IPv4 and IPv6 change the type of attacks
which IPv6 networks are likely to see.
3.1 End-to-End
As stated in [3], section 6, there is a problem with end-to-end
communications, which means that every host should be reachable from
any other host, included the ones from "outside". This is a problem
in the network-based security model.
This kind of communications provides the required framework for
further innovation, where technologies like P2P and GRID (among
others) can be widely spread with no problems.
Vives, et al. Expires April 24, 2005 [Page 10]
Internet-Draft IPv6 Security Problem Statement October 2004
In [3] some possible solutions are outlined, one of them being a host
firewall.
3.2 IPsec-encrypted ESP-traffic in transport mode
As stated in [3], section 5, there is a problem with the IPv6
encrypted traffic (IPsec ESP mechanism in transport mode, for
example) and the network-based security model.
The idea is that a host inside the network can establish an encrypted
communication channel with other host outside of the network, for
example. A middlebox (for example the perimeter firewall) won't be
able to inspect the content of such a communication..
In [3] some possible solutions are outlined, one of them being a host
firewall.
3.3 Mobility
In parallel to the increase in the number of devices, IPv6
facilitates that those devices are "mobile", that is, can easily move
from one network to another using Mobile IPv6 or just disconnecting
from one and connecting to another.
Because of the amount of addresses available and the facilities given
by Autoconfiguration mechanisms together with the mentioned rise of
the number of IP devices, this kind of behavior should be taken into
account by the security administrator, as these devices will be
connected to networks where they have no control and consequently, no
responsibility.
A possible solution for these devices is the use of host based
security, enabled in every network it is connected. The policies and
mechanisms should be described elsewhere.
3.4 Addresses
Regarding the addresses in IPv6 must be taken into account that:
o The amount of addresses is much bigger for a given network.
o Each host will have more than one address which are probably
globally routable.
o An IPv6 node can use randomly generated addresses [4].
That means:
Vives, et al. Expires April 24, 2005 [Page 11]
Internet-Draft IPv6 Security Problem Statement October 2004
o To scan a given network whole range of addresses and ports will
take a really big effort [5]. It would be easier to do that by
sniffing a LAN segment looking for existent addresses.
o The common way of identifying a host by means of its IP address
will be more difficult to use.
o If a host uses randomly generated addresses [4], it could be
problematic to identify a host using its IP address for security
policy matching purposes.
Regarding the scan of addresses, [5] demonstrates that the "brute
force" scanning would make no sense for an IPv6 address range,
typically a minimum of /64.
A host based security scheme would protect the other hosts from the
compromised one.
The idea behind all this is that the new IPv6 address scheme and
mechanisms will somehow protect from existent attack techniques but
we can be sure that they will adapt themselves to the new scheme and
we have to act consequently being prepared.
3.4.1 Link-local addresses
In IPv6 we have got the link-local addresses that allows a host that
connects to a network to have IP connectivity without any external
help.
Even if this is quite useful, also represents a security problem
because allows a host to attack the network it is connected to. This
must be taken into account by the security administrator.
As a guideline, we should not simply rely on trusting by default
those sessions which are from link local addresses. It is better to
restrict to use link local address to some fundamental services,
until the host is trusted.
3.4.2 New Multicast addresses
In IPv6 new multicast addresses are defined than identifies resources
in a well known manner. This can enable an enemy to locate and
attack those key resources with no need of a time consuming address
search.
This kind of addresses have an scope field. This means that many of
such services would have either link-local scope. Note that these
addresses are used for the protocol itself, for example in the
Vives, et al. Expires April 24, 2005 [Page 12]
Internet-Draft IPv6 Security Problem Statement October 2004
autoconfiguration process. This kind of addresses must be taken into
account by a security solution that addresses inside attacks.
There are also global-scope addresses that must be published to the
outside. This kind of services must be protected and monitored by a
security solution that addresses outside attacks.
For an updated IPv6 multicast addresses list see [6].
3.5 Neighbor Discovery Weakness
As said above, one of the assumptions of the host-based security
model is that all hosts in the network are non trusted, the possible
threats coming from the same LAN segment must be taken into account,
in this case the ones coming from Neighbor Discovery (ND) [7][8].
Note that this is not possible within the perimeter security model,
although some detection mechanism could be implemented, nothing can
be done to protect the hosts.
There are some ways to interfere in the normal behavior of the
autoconfiguration process, causing redirection of traffic and/or DoS
(Denial of Service).
Special attention must be put on Router Advertisement (RA), Router
Solicitation (RS), Neighbor Solicitation (NS), Neighbor Advertisement
(NA) and Redirect messages. See [9] for a detailed explanation of
possible threats.
Using host firewalls and/or IDS (Intrusion Detection Systems) for
protecting hosts against these threats is very likely a wrong
approach, as that would basically imply reinventing SEND; it is
better to use SEND instead [10][11].
There is no easy protection against these threats as the ND features
(e.g., using link-local or the unspecified addresses) are needed
before a host can authorize itself to the other network components.
3.6 Routing Header
TBD.
3.7 Home Address Option
TBD.
Vives, et al. Expires April 24, 2005 [Page 13]
Internet-Draft IPv6 Security Problem Statement October 2004
3.8 Embedded Devices
With the deployment of IPv6 we can expect the avenue of a big amount
of new IPv6-enabled devices with few resources, low computing
capacity, even low battery capacity. In some cases, this kind of
devices will not be able even to perform the minimum set of functions
required by the Host-based Security Model.
This should be taken into account when the security requirements are
outlined and by the proposed solutions.
4. Other Issues
Further elaboration is required (TBD) on:
o Malicious users: We can't protect the network from malicious users
that have physical access to network hosts in the protected
network. The objective is to minimize the danger they can cause.
o In the host-based security, the host that stores and distributes
the security policies seems to be the best option to be the one
that acts as IDS.
5. Security Considerations
This document is concerned entirely with security.
6. Acknowledgements
The authors would like to acknowledge the inputs of Brian Carpenter,
Satoshi Kondo and the European Commission support in the co-funding
of the Euro6IX project, where this work is being developed.
7. References
7.1 Normative References
7.2 Informative References
[1] "IETF midcom WG",
<http://www.ietf.org/html.charters/midcom-charter.html>.
[2] Bellovin, S., "Distributed Firewalls", November 1999,
<http://www.research.att.com/~smb/papers/distfw.pdf>.
[3] Savola, P., "Firewalling Considerations for IPv6",
draft-savola-v6ops-firewalling-02 (work in progress), October
Vives, et al. Expires April 24, 2005 [Page 14]
Internet-Draft IPv6 Security Problem Statement October 2004
2003.
[4] Narten, T. and R. Draves, "Privacy Extensions for Stateless
Address Autoconfiguration in IPv6", RFC 3041, January 2001.
[5] Chown, T., "IPv6 Implications for TCP/UDP Port Scanning",
draft-chown-v6ops-port-scanning-implications-01 (work in
progress), July 2004.
[6] "IANA's IPv6 Multicast Addresses List",
<http://www.iana.org/assignments/ipv6-multicast-addresses>.
[7] Narten, T., Nordmark, E. and W. Simpson, "Neighbor Discovery
for IP Version 6 (IPv6)", RFC 2461, December 1998.
[8] Thomson, S. and T. Narten, "IPv6 Stateless Address
Autoconfiguration", RFC 2462, December 1998.
[9] Nikander, P., "IPv6 Neighbor Discovery trust models and
threats", draft-ietf-send-psreq-04 (work in progress), October
2003.
[10] Arkko, J., Kempf, J., Sommerfeld, B., Zill, B. and P. Nikander,
"SEcure Neighbor Discovery (SEND)", draft-ietf-send-ndopt-06
(work in progress), July 2004.
[11] Aura, T., "Cryptographically Generated Addresses (CGA)",
draft-ietf-send-cga-06 (work in progress), April 2004.
Authors' Addresses
Alvaro Vives Martinez
Consulintel
San Jose Artesano, 1
Alcobendas - Madrid
E-28108 - Spain
Phone: +34 91 151 81 99
Fax: +34 91 151 81 98
EMail: alvaro.vives@consulintel.es
Vives, et al. Expires April 24, 2005 [Page 15]
Internet-Draft IPv6 Security Problem Statement October 2004
Jordi Palet Martinez
Consulintel
San Jose Artesano, 1
Alcobendas - Madrid
E-28108 - Spain
Phone: +34 91 151 81 99
Fax: +34 91 151 81 98
EMail: jordi.palet@consulintel.es
Pekka Savola
CSC/FUNET
Espoo
Finland
EMail: psavola@funet.fi
Vives, et al. Expires April 24, 2005 [Page 16]
Internet-Draft IPv6 Security Problem Statement October 2004
Intellectual Property Statement
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be claimed to
pertain to the implementation or use of the technology described in
this document or the extent to which any license under such rights
might or might not be available; nor does it represent that it has
made any independent effort to identify any such rights. Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to implement
this standard. Please address the information to the IETF at
ietf-ipr@ietf.org.
Disclaimer of Validity
This document and the information contained herein are provided on an
"AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE REPRESENTS
OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND THE INTERNET
ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE
INFORMATION HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Copyright Statement
Copyright (C) The Internet Society (2004). This document is subject
to the rights, licenses and restrictions contained in BCP 78, and
except as set forth therein, the authors retain all their rights.
Acknowledgment
Funding for the RFC Editor function is currently provided by the
Internet Society.
Vives, et al. Expires April 24, 2005 [Page 17]