IPv6 Security: Transition from I Pv 4 to I Pv 6

painlosososΛογισμικό & κατασκευή λογ/κού

30 Ιουν 2012 (πριν από 4 χρόνια και 11 μήνες)

324 εμφανίσεις

IPv6 Security: Transition from IPv4 to IPv6
Daniël van der Steeg
University of Twente
P.O. Box 217, 7500AE Enschede
The Netherlands
d.p.m.h.vandersteeg@student.utwente.nl




ABSTRACT

Because of the transition from IPv4 to IPv6 and certain
differences between both IP versions, there are quite some
possible security issues. In this paper we try to find how the
security of networks is affected by this transition. For this
purpose we made a questionnaire to inquire about the general
status of the transition in companies which provide services
connected to networks (ISPs, security providers, etcetera).
Secondly we used practical tests on production networks in
order to get an idea about the IPv6 networks in the wild. The
questionnaire shows that IPsec is not used more with IPv6
than with IPv4. Furthermore it shows that IPv6 and IPv4 are
considered to be equally secure, while the IPv6 appliances
used still handle IPv6 in software a lot. The practical tests
supported the hypothesis that IPv6 networks are less secure
than IPv4 networks. Even though the scale of this study is
limited, the results are important because they show that the
security of IPv6 still has a long way to go.
Keywords
IPv6, transition, security, network, IPv4
1. INTRODUCTION
Current networks are mostly based on the IP protocol: the
entire internet relies on it for the propagation of information.
The currently most used version is IPv4, but this version only
allows for approximately 4.3 billion addresses. More than a
decade ago, the conclusion was drawn that we would
eventually run out of addresses because with allocating
addresses, a lot of them were wasted: a block of 65 thousand
addresses would be assigned to one company with far less
computers. Consequently, in December 1995 the new version
was specified: IPv6 [6]. However, back then no one was
interested in the new version of the Internet Protocol because
most people did not see an immediate need for it. Only in the
last few years the IPv6 usage started to grow more rapidly.
With the rapidly growing number of machines connected to
the Internet, the total number of IPv4 addresses is starting to
look a bit bleak. Now that also the worldwide IPv4 address
pool is depleted [8], network providers are urged more and
more to find solutions for the shrinking amount of addresses,
of which one is to migrate to IPv6. Some of the Dutch ISPs
have already moved to IPv6 like SURFnet [11] and XS4ALL.
Some major other service providers, like Google and
Facebook, are integrating IPv6 into their networks and will be
testing this worldwide on World IPv6 Day[7]. On that day
major organizations will offer their content to world over IPv6
for 24 hours as a test flight.
1.1 Focus of This Paper
When migrating to a different technology certain security
issues may arise. The mandatory implementation of IPsec in
IPv6 makes some parties think it will be more secure [9]. But
this is not necessarily the case: implementing it does not mean
using it. Even when using IPsec, there are still attacks possible
against such networks [13]. Aside from the mandatory
implementation of IPsec, there are other differences between
both protocols. An example of this is that the larger address
space makes it harder for someone to do reconnaissance
attacks [1, 12]. It would take quite some time to scan 2


addresses per network. Another difference between both IP
versions is that IPv6 heavily relies on ICMPv6 [1] which
makes it harder to setup firewalls to be safe without removing
functionality from IPv6, but allowing everything through the
firewall can pose a threat. All of these differences might have
an impact on security, because they might not be known or
fully understood by network administrators or they do not see
the immediate threat because IPv6 is not widely used yet.
Something else that might have an impact on the security of
IPv6 might be that network security devices are not capable of
handling IPv6 properly yet, resulting in only partially secured
networks. With all those possibilities to have a less secure
network, the question which arises is whether IPv6 is more or
less secure than IPv4, in practice.
This paper will try to answer the question “How does the
transition from IPv4 to IPv6 affect the security of networks?”
To answer this question, first the following questions will be
answered:
￿ What is the state of the transition towards IPv6 in
networks?
￿ What is the state of the transition towards IPv6 in
security products?
￿ What is the security state of the IPv6 enabled
networks and how does this differ from IPv4?
The results will consist of an overview of the current state of
the transition to IPv6 ‘in the wild’. It will be used to test the
following hypothesis: “IPv6 is in practice not a more secure
protocol than IPv4.”
In order to answer the above mentioned questions first an
analysis of the available literature is made to get an overview
of differences between the IP versions and possible security
problems. This is discussed in the next section. Secondly a
survey was conducted among companies which work a lot
with networks, such as ISPs and hosting providers. This
should give an idea about how IPv6 is used and implemented

Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that co
pies are
not made or distributed for profit or commercial advantage and that
copies bear this notice and the full citation on the first page. To copy
otherwise, or republish, to post on servers or to redistribute to lists,
requires prior specific permission and/or a fee.
15
th
Twente Student Conference on IT, June 20
th
, 2011, Enschede, The
Netherlands.
Copyright 2011, University of Twente, Faculty of Electrical Engineering,
Mathematics and Computer Science.
in practice. Lastly the network of the University of Twente
has been tested in order to find differences in security between
IPv4 and IPv6.
The questionnaire and the practical tests are described in
section 3 after which the results are described. Finally there
will be some discussion about the results and a conclusion
will be drawn.
2. STATE OF THE ART
First some functions of IPv6 that are mentioned throughout
this paper will be explained and then some research into IPv6
will be discussed.
2.1 IPv6 Functions
IPv6 makes use of State-Less Address Auto Configuration
(SLAAC) to obtain an address if there is no DHCPv6 server
on the network. This configuration normally takes the link-
local prefix fe80::/64 as network prefix while the lower 64
bits are populated by expanding the MAC address from 48
bits to 64 bits using modified EUI-64. The reason for taking
the MAC address is that they should be unique and thus there
should not be collisions. However this is not always the case,
so in order to be sure there is no collision it uses ICMPv6’s
Duplicate Address Detection, which asks to the nodes on the
network if this address already exists using a Neighbor
Solicitation packet. If it does exist, it stops the auto-
configuration and needs manual configuration, otherwise it
uses this address. For global addresses it uses the network
prefix the router advertised to the machine instead of the link-
local prefix. This process has been depicted in figure 1.

Figure 1. Stateless Address Auto Configuration
Since the MAC address is used, it would be easy to follow a
roaming client since the last 64bits will always be the same
(assuming there are no collisions and the user does not change
the MAC address). To prevent this, there are Privacy
Extensions which randomize the IPv6 address each time the
machine needs a new one. However, it makes it hard to trace
back certain traffic on a network to a specific client since the
last 64 bits cannot be reversed to a MAC address and thus a
certain computer (assuming the MAC is not changed
temporarily).
In order to find a router, a node can send a Router Solicitation
to which a router will reply with a Router Advertisement
(RA). In addition to answering solicitations, the router also
can send this advertisement periodically, for example to
update certain information. This Router Advertisement
contains among other information the network prefix which
the nodes must use for their IPv6 addresses, the lifetime of the
router which indicates the period in which the nodes may
consider the router valid and a preference level, which can be
used to order the preference of the router in the nodes. This
information is used by the nodes in the network to create an
address and to build the routing table.
2.2 Related Works
There has already been a lot of research into the theoretical
differences between IPv4 and IPv6. For example, Caicedo and
Joshi [1] show some of the security challenges IPv6 poses
such as potential problems with the State-Less Address Auto
Configuration feature of IPv6, which could facilitate a denial
of service attack. This could happen for example if the
attacker prevents the target from obtaining an address through
SLAAC by pretending the pending address is already in use.
Other problems which are discussed are attacks using routing
headers which could help an attacker reach a node which
would normally drop such traffic and multicast-based attacks
which might use the All Router address. Furthermore, they
discuss that filtering IPv6 packets will be burdensome due to
the amount of possible IPv6 headers and that the place
ICMPv6 takes in IPv6 makes it hard to define firewall rules
for that as well. They conclude with a short discussion about
transition issues.
In [3] Choudhary elaborates on [1] and discusses another
attack against IPv6 nodes: a denial of service attack against
security devices which need the whole packet for deep packet
inspection. By not sending the last fragment their buffers will
be fully allocated after a while. IPv6 fixed this issue for
routers by not allowing them to reconstruct the packet, but
nodes that do are still vulnerable. Also a way of evading
detection of attacks is suggested: one could use chains of
routing headers to make it hard for firewalls to assess the
traffic. Those chains are compliant with the IPv6 specification
and thus cannot be dropped as a precaution. Some types of
attacks using SLAAC and ICMPv6 are mentioned including:
an attacker might insert himself as router and sniff all the
traffic or just drop it, an attacker can send out router
advertisements pretending to be the real router with lifetime
set to zero causing a denial of service, an attacker can send
messages to nonexistent nodes causing the routers to neighbor
detect them which will fail and the router therefore wastes
resources. In addition it is possible to send bad network
prefixes which can cause nodes to be unreachable.
Carp et al. [2] implemented and evaluated different attacks
against IPv6 in local area networks and discuss mitigation of
such attacks. They test active and passive reconnaissance
attacks respectively using Neighbor Solicitation and sniffing
for ICMPv6 traffic. In order to reduce the number of
addresses to scan, a few bits were ‘frozen’. This is possible
when the addresses are obtained by SLAAC and a part of the
MAC address is known (e.g. the manufacturer specific part).
Also some reconnaissance attack optimizations are proposed,
tested and evaluated. These include overflowing the MAC
table of a switch so it becomes a hub and pretending to be a
router or root bridge so nodes contact the attacker themselves.
Furthermore they successfully test a denial of service attack
which prevents a node from assigning an IPv6 address to itself
by spoofing messages from potential IPv6 addresses of the
target and thus preventing the target from auto configuring an
address.
Yang et al. [13] describe that IPsec is still vulnerable to some
attacks like attacks against passwords and the Internet Key
Exchange infrastructure. It also introduces a possible new
problem: since encrypting uses quite some resources IPsec
might open up new opportunities for denial of service attacks.
3. APPROACH
3.1 Questionnaire
In order to get an understanding of the status of IPv6 in the
wild, it is necessary to get in contact with companies and
organizations that use IPv6. There were two possibilities,
either a questionnaire or interviews.
The questionnaire was chosen in this case for several reasons.
It is more distant and quicker than interviews which might
make it easier for companies to cooperate. Furthermore a
questionnaire makes it easier and faster to analyze data if most
of the questions are closed ended.
The questionnaire consists of 34 questions in total; divided in
such a way that not all questions will be reached. The reason
for this setup is that not all of the questions are needed for all
the respondents. Some of the questions also are not necessary
if the respondent answers a certain question in a certain way.
The questions are made up of 26 closed questions, of which
25 questions are made to find actual information. These
questions include rating scales (How secure do you consider
your IPv6 network compared to your IPv4 network?),
multiple choice questions with both multiple answers (What
type of security measures do you employ for IPv6?) and with
only one possible answer (When did you introduce IPv6?).
The first question is a demographic question in order to find
out how much the respondent might know of the network. The
other 8 questions, which are open ended questions, are used to
find the motivation of a previously given answer, or to let the
respondent add extra information to a given answer.
As targets for the questionnaire different companies were
chose which all have some business with the internet. There
were mainly three types: internet service providers (mostly
Dutch) such as XS4ALL, UPC and KPN and some hosting
providers, big companies which provide their services through
the internet such as Google, Hyves etcetera. and finally
companies which sell network security appliances, such as
Cisco, Symantec and Checkpoint. Aside from those
companies, the questionnaire was also send to the Dutch IPv6
Taskforce mailing list, because a lot of companies are
subscribed to it and the people reading it are mostly technical
so it is a good point of contact.
The full questionnaire can be found in Appendix A.
3.2 Practical Test
Aside from a questionnaire a practical test was carried out on
the network of the University of Twente, because this can be
used to describe how IPv6 is used in practice, as opposed to
theory, and in a real environment, not a lab environment. The
tests consist of three parts: testing IPsec in the network,
testing different ways of identifying nodes on an IPv6
network and lastly testing the consistency of the security
settings of the network between IPv4 and IPv6 and testing
some other security related issues.
3.2.1 Testing IPsec
With the compulsory implementation of IPsec in IPv6, one
might wonder if this is actually used throughout the networks,
since there are quite some alternatives available, for example
SSL. Beforehand the network administrators already noted
that the services of the university were not using IPsec,
consequently this test is meant to see if the routers and
firewalls can properly handle the IPsec traffic.
The environment of this experiment is as follows: two
computers running Linux, both wired to the network. Both
computers have IPsec-Tools installed, which is a port of
KAME's IPsec utilities to Linux-2.6. One of the computers is
running Wireshark as well, in order to capture the data that is
send between them. This machine was also running a basic
Apache server as a means of testing the connectivity between
the two computers. The IPsec settings were set to require a
connection with Encapsulating Security Payload using 3DES-
CBC encryption. Once these settings were in place on both
machines, the webpage was loaded to see if the IPsec traffic
was routed and this traffic was logged with Wireshark.
3.2.2 Identifying nodes on IPv6 network
IPv6 comes with 128-bit addresses, of which usually 64 are
the network prefix and the other 64 bits are left to identify a
node in the network. Due to this enormous increase in
possible addresses, a ping scan to map a network will not
work as well with IPv6 as it did with IPv4. Scanning one
network would require probing 2

addresses, which would
take too long. Even if one would optimize this scanning (e.g.
focusing on a specific network card manufacturer as described
in [2]) and thus restricting the scan a lot, this would be an
address space of 2

addresses. In order to find nodes
attackers will eventually find other ways, e.g. using Neighbor
Detection for this, using multicast addresses for this (e.g.
FF05::1 for all nodes on a site) or querying the local DNS
servers for information about the network. The goal of this
experiment is to test other methods of gaining knowledge
about the network. The experiment consists of two separate
tests, the first one being trying to use DNS servers to find
information, the second being using multicast addresses to
find nodes on the network.
The environment for this experiment consisted of a computer
running Linux wired to the network. For the first experiment
the program dig was used to query the DNS servers. First the
IPv6 address prefix of the machine used was inverted and
given as parameter to the dig program yielding the result
below.
dig AXFR 8.0.9.1.0.1.6.0.1.0.0.2.ip6.arpa
AXFR is the transaction type which is mostly used by
administrators to replicate the contents of a DNS, in order to
move it to a slave DNS.
For the second experiment the ping program was used to send
a ping to the Link-Local All-Nodes multicast address FF02::1.
The ping program was also used to receive all the responses.
In order to secure a network against unwanted reconnaissance
attacks, networks should be segmented properly and DNS
servers and routers should be properly secured to prevent easy
access to information about the network. Allowing anyone to
extract all the information from the DNS might give an
attacker a good idea about where the interesting machines are.
3.2.3 Testing security settings
In the transition period a lot of networks will have both IPv4
and IPv6 enabled, whether it be dual-stack, tunneling or
translation. However, using both protocols also requires
securing both protocols. Problems may occur when IPv6 is
enabled but only IPv4 is protected, since this would allow an
attacker to still attack a node in the network if he manages to
map the IPv4 address to an IPv6 address which he might do
using for example DNS records. This experiment is used to
get an idea about how well secured a network which operates
on both IP versions in practice is. It consists of testing firewall
settings by port scanning a certain machine from inside the
network as well from outside the network and testing an
aimed denial of service attack using the ICMPv6 Router
Advertisement packet.
The environment of these experiments consists of two
computers wired to the university’s network and one machine
outside of the university’s network (see figure 2).

Figure 2. Security Settings Experiment Setup
The reason for this setup is that usually only traffic crossing
the border routers is matched against an Access Control List
(ACL). The first computer was actively used for the tests
within the network and was running Nmap and Scapy. Nmap
was used to scan the open ports on the second machine. Scapy
was used to test the denial of service attack using RA packets.
The second machine was running an Apache server, SSH and
had ports 25, 12345 and 31337 open using the netcat program.
These ports were chosen because according to [10] they
should be blocked on the university’s routers. The third
machine was only running Nmap to do the port scan from
outside the university’s network.
The first experiment, the port scans, was set up as stated
above. First machine 1 scanned all ports of machine 2 over
IPv4, then over IPv6. Secondly machine 3 scanned all ports of
machine 2 over IPv4 and then over IPv6.
The testing of the DoS attack was carried out by machine 1, in
Scapy an IPv6 packet was made with the IPv6 address of
machine 2 as destination encapsulating an ICMPv6 Router
Advertisement packet with the routerlifetime option set to
zero. This should result in the node dropping routes to that
router because the router is not valid anymore. Consequently
this machine cannot connect to the internet anymore because
it thinks there is no router.
For this part of the experiment machine 2 was also running
Wireshark in order to find out if the packets were actually
received.
After testing with only the address of machine 2 as
destination, the usual destination was also used: ff02::1, the
well-known link-local all nodes address.
4. RESULTS
In this section the results of the survey and the experiments
will be given subsequently. The results will be evaluated and
discussed in section 5.
4.1 Questionnaire
Of the 25 companies contacted (excluding the IPv6 Taskforce
mailing list) roughly 11 responses were useful, the others
were only partially filled in or just skipped through. Only the
more interesting answers are discussed here.
In the first question of the questionnaire, the respondents were
asked to indicate their function within the company they work
for. Half the respondents indicated they were somehow
involved with the company's network. The other half of the
respondents indicated they occupied a management function.
The second question inquired about what type of company the
respondent works for. Here 55% responded that their
company provided internet services (i.e. ISPs and hosting
providers etc.). The other 45% answered Other which was
defined to be a university and a consultancy agency among
others. The option that the company provided security
services was not used.
The next question asked which protocols were supported by
the services their company provided. All the respondents
indicated that IPv4 was supported while 91% responded IPv6
is also supported by their services and 82% responded that
also the backbone of their network supports IP version 6. One
reason given for not using IPv6 was that it is too expensive.
When asked what the reason is to employ IPv6 in their
networks 80% of the people who stated to use IPv6, answered
it was out of anticipation of future need. The rest has a
different reason for using IPv6.
The following question inquired about when IPv6 was
introduced, the results can be found in figure 3.

Figure 3. Chart on when IPv6 was introduced in the
respondents networks.
In the tenth question in the questionnaire, the respondents
were asked what the status is of IPv6 in their services. Of the
respondents who answered to use IPv6, 60% answered IPv6
was deployment ready, i.e. it is already deployed or it is ready
to be deployed. The remaining 40% answered their IPv6 was
still in a testing stage.
All the respondents answered they used security measures in
their network both in IPv4 and in IPv6. When asked how
secure they considered their network, 82% answered they
considered their IPv4 network secure, while 18% answered
they considered it not secure nor unsecure. For IPv6 the
percentages are respectively 80% and 20%.
Table 1. Results of questionnaire question: ‘What security
is used in your network?’

IPv4
IPv6
Firewall
88,9%
87,5%
Intrusion Detection System
44,4%
25%
Intrusion Prevention System
22,2%
0%
IPsec
33,3%
37,5%
Other
55,6%
50%
The results of the question which security measures they use
for IPv4 and IPv6 can be found in table 1. The percentages are
broken down by the amount of respondents who answered to
use the IP version in question.
To the question why the IPsec is not being used in the
network, 40% responded it was unnecessary and the others
were equally divided over ‘Too difficult to set up’, ‘Not much
more secure’ and ‘Other reason’.
The 22
nd
question of the questionnaire asked the respondents
about whether the settings in their security appliances are
consistent between both IP versions. 63% of the respondents
said this was the case, while 37% stated that this wasn’t the
case. One reason for this was the difference between both
versions.
Next they were asked if their network had already been target
of any attacks. The respondents answered as shown in figure
4. The people who stated their network had already been
attacked, were then asked what type of attacks their network
had been target of, to which most of them replied with higher
layer protocols such as HTTP and SSH.

Figure 4. Result of question ‘Has your IPv6 network
already been target of attacks?’
The security appliances used by the respondents that use IPv6,
mainly support it in software only (57%). 14% of the
respondents have IPv6 support in hardware and the rest have
IPv6 supported in both.
Finally 89% of the contestants consider IPv4 and IPv6 to be
equally safe in general, while the remaining 11% considers
IPv6 to be less secure than IPv4.
4.2 Practical Test
4.2.1 Testing IPsec
As stated in the description of this experiment, the services of
the university are not using IPsec, so this is focused on finding
whether or not the routers could properly handle IPsec traffic.
After setting up the experiment environment, the webpage on
the host machine was accessed successfully. Wireshark also
reported that the protocol used was ESP and the contents of
the packets were unreadable proofing that the encryption
worked.
4.2.2 Identifying nodes on IPv6 network
In order to find nodes on an IPv6 network, two experiments
have been carried out. The first one involved querying the
local DNS in order to find registered domain names which are
linked to an IPv6 address. Using the ‘dig’ program as stated in
the description of this experiment, resulted in 1733 IP
addresses being returned in approximately 72 milliseconds
(albeit in reverse). The IPv4 variant returned a very long list
of secondary DNS servers which hold the data. These servers
have mapped nearly every IPv4 address to a domain name and
therefore the total amount of entries will be quite a lot.
The second part tested the combination of the multicast
address FF02::1 and the ping program in order to find nodes
on a network. This test resulted after approximately one
second in a set of 309 active link local addresses which can be
transformed to worldwide addresses if the network prefix is
known.
4.2.3 Testing security settings
The first experiment was a port scan from inside the
university’s network to another computer inside the same
network. The machine which was the target of the scan had a
few open ports, namely 22, 25, 80, 9876, 12345 and 31337.
All of these ports were found to be open by Nmap over both
IP versions, so all the traffic was routed without interference
to the target machine.
The second experiment consisted of a port scan from outside
the university’s network, to a machine inside it. Again ports
22, 25, 80, 9876, 12345 and 31337 were opened on the target
machine, but only ports 22, 80, 12345 and 31337 were found
open when using IPv4, port 25 and other ports were listed as
filtered. The same scan over IPv6 returned only the six open
ports. These results can be found in Table 2.
Table 2. Results of border crossing portscan from outside
the University’s network to inside it.
PORT
IPv4
IPv6
SERVICE
22/tcp
open
open
ssh
25/tcp
filtered
open
smtp
42/tcp
filtered
-
nameserver
80/tcp
open
open
http
135/tcp
filtered
-
msrpc
136/tcp
filtered
-
profile
137/tcp
filtered
-
netbios-ns
138/tcp
filtered
-
netbios-dgm
139/tcp
filtered
-
netbios-ssn
161/tcp
filtered
-
snmp
162/tcp
filtered
-
snmptrap
445/tcp
filtered
-
microsoft-ds
1434/tcp
filtered
-
ms-sql-m
5555/tcp
filtered
-
freeciv
9876/tcp
open
open
http/ssl
12345/tcp
open
open
netbus?
31337/tcp
open
open
Elite?

The last experiment of the practical tests consists of trying to
disable a machine by sending him false information about a
router. The goal is to get the target to delete the router from
his route table, resulting in it not being able to send anything
to the internet. The result of this experiment, both aimed
(sending only to the target) and spread (sending it to FF02::1)
failed. The routing table of the target machine did not change.
5. DISCUSSION
In this section the previous described results will be
interpreted. First some general remarks regarding the research
will be given in order to identify certain weak points. The first
of those remarks is that this research only gives information
about what the status of IPv6 is at the moment of the research:
it is a snapshot. However, it might for example be the start of
a periodical research which strives to map the progress of the
deployment of IPv6.
5.1 Questionnaire
Due to time and resource constraints, the number of
respondents of the questionnaire it not really high and
therefore the results do not have any statistical significance.
This does not mean there are no trends to find in the results.
First of all, the number of people who responded to support
IPv6 seems to be quite high. There are different ways in
which IPv6 connectivity can be measured. One of them is the
amount of users actually using IPv6 to connect to the internet.
Google for one is measuring the amount of people that use
IPv6 to use their services [4]. According to these
measurements currently about 0.33% of the people using
Google use IPv6 to do so. However, since Google’s IPv6 is
only enables for certain whitelisted networks, this may not be
totally representative. A different kind of measure is the
amount of Autonomous Systems (AS) that announce IPv6
address blocks compared to the amount announcing IPv4
blocks. Since ASs are roughly speaking the same as an ISP
this would indicate how many networks are IPv6 capable.
According to the numbers collected by [5] this means that
2.93% of the networks are IPv6 capable. However, this too
might not be representative because some networks might
announce IPv6 while not the whole network is IPv6 capable.
Reasons for this enormous difference between our results and
above mentioned numbers might be that people whose
company does not support IPv6 do not think their contribution
is of any worth, or the group of companies (including the IPv6
taskforce) is not really representative. This number of IPv6
capable respondents is however useful when analyzing IPv6
security.
According to the results, most people think their networks are
quite safe, and when asked what types of security measures
they use something interesting shows: in order to secure IPv6
less measures are taken, while most respondents consider both
‘parts’ of the network to be equally secure. The amount of
people who stated to use IPsec in IPv6 is remarkable as well,
it was supposed to be the security feature of IPv6 but it’s
usage is not significantly higher than with IPv4. Most
respondents believe that they do not need it, albeit for
different reasons. In addition most respondents seem to think
IPv4 and IPv6 are generally equally safe while the respondent
who answered IPv6 was less secure, also did not use IPv6 in
the network.
Something also noteworthy is that the security appliances the
respondents use, mostly support IPv6 in software. A problem
might occur when the firewall has so much traffic to process;
it cannot handle it and might end up exhibiting unwanted
behavior.
5.2 Practical Tests
The results of these tests show that the University of Twente
does not use IPsec itself, but that the hardware can route it
without problem, as expected.
As for the experiment regarding the identification of nodes on
an IPv6 network, there is one thing that should be noted: a
ping scan using Neighbor Detection was not experimented
with, because this would flood the routers neighbor tables
which might cause them to crash and consequently disrupt the
network as described in [3].
It can be observed from the results that aside from a tedious
ping sweep, there are other ways to find online machines in a
network. One conclusion might be that network administrators
need to secure the network better against such reconnaissance
attacks, because they might become a lot more common.
The results from the security settings experiment show that
there is quite a difference between the security of IPv4 and
IPv6. The IPv4 network is protected from the outside world
using an Access Control List, but the IPv6 network is not
protected at all according to the results. The ports with a dash
in Table 2 were not listed by nmap because they are not
filtered (as with IPv4) nor open on the target machine and can
therefore be considered open. Two things are interesting to
note: the samba port is not filtered either, leaving it open to
the world and the list in [10] does not match the list of filtered
ports.
Even when taking into account that according to ICTS there
are no services from the university available over IPv6, it is a
remarkable result because it still leaves open regular nodes on
the network. Indirectly this can still cause harm.
It should be noted however that IPv6 it not totally forgotten.
Getting an IPv6 address using SLAAC with privacy
extensions would make it nearly impossible to trace anything
back as described in section 2.1. However, by making
backups of the neighbor tables of the routers they manage to
create a link from an IPv6 address back to a MAC address.
This allows them to trace everything back to a certain person
if this should be necessary.
We could not trace why the DoS attack failed since the
packets did arrive at the target machine. Even when copying
nearly all the fields from the Router Advertisements of the
actual routers on the network, the results stayed the same. The
theory makes for a good addition to the DoS attacks described
in [1-3].
6. CONCLUSIONS
The results of the questionnaire showed that a lot of networks
have already started to migrate to IPv6. This is only the start,
slowly more and more networks will reach the state were IPv6
is usable because there is a lot of pressure due to the limited
amount size of the IPv4 address space. The security products
do support IPv6 as well, but not optimally yet. The
questionnaire shows that IPv6 is handled mostly in software
which might cause problems. Furthermore, most respondents
answered they do not use IPsec, which was very important in
the early protocol specification. It also shows that IPv6
networks are under attack, and thus must be secured. Results
of the practical tests show that the security of IPv6 networks
is not optimal compared to IPv4, which might be caused by
the lack of enthusiasm for IPv6.
For now we can conclude that networks running IPv6 are at
least not safer than networks running IPv4. The security of
networks mainly depends on their deployment, and since IPv6
is still young, networks running IPv6 will most likely become
more robust over time.
7. FUTURE WORK
There is a good possibility for interesting further research.
Part of this paper was about trying to get an idea of how IPv6
networks are deployed and secured in the real world and not
only in test environments. Since the lack of IPv4 addresses is
pushing network administrators towards IPv6, this area is ever
changing. It would be interesting to do the same research
periodically to monitor the progress of the transition towards
IPv6. In addition, results from such research might be used to
motivate people to speed up the transition and urge them to
secure their networks, especially in a world in which the
internet is still becoming a substantial part of everyday life.
When repeating this research, there are a few enhancements
possible. In the questionnaire was asked how much of the
traffic flowing through the respondent’s network is IPv6,
making the ranges in the answers smaller will give a better
idea about how much traffic it actually is. Moreover, the
number of respondents should be higher to get statistically
significant results. Finally the practical tests might be
expanded to get more coverage and therefore get a better
understanding where the security issues lie. This should
include testing different networks and more possible security
issues.
8. ACKNOWLEGDEMENTS
The author would like to thank Frank Kargl for valuable
comments given during the research period and also ICTS for
allowing their network to be tested and for the information
about their network.
9. REFERENCES
[1] Caicedo, C. E., Joshi, J. B. D. and Tuladhar, S. R. IPv6
Security Challenges. Computer, 42, 2 (Feb. 2009), 36-42.
[2] Carp, A., Soare, A. and Rughiniş, R. Practical analysis of
IPv6 security auditing methods. In Proceedings of the
9th RoEduNet IEEE International Conference,
RoEduNet 2010 (Sibiu, Romania, June 24-26, 2010).
IEEE, 2010.
[3] Choudhary, A. R. In-depth analysis of IPv6 security
posture. In CollaboarteCom 2009 (Washington, DC,
USA, Nov. 11-14, 2009). IEEE, 2009.
[4] Google. IPv6 Statistics.
http://www.google.com/intl/en/ipv6/statistics/. Visited on
June 1st, 2011.
[5] Huston, G. IPv6: IPv6 / IPv4 Comparative Statistics.
http://bgp.potaroo.net/v6/v6rpt.html. Visited on June 1
st
,
2011.
[6] IETF. RFC 1883 Internet Protocol, Version 6 (IPv6)
Specification. December 1995.
[7] ISOC World IPv6 Day.
http://isoc.org/wp/worldipv6day/. Visited on Febuary
28th 2011.
[8] NRO. Free Pool of IPv4 Address Space Depleted.
http://www.nro.net/news/ipv4-free-pool-depleted.
Number Resource Organization, Febuary 2011.
[9] Rowe, B. and Gallaher, M. Could IPv6 Improve Network
Security-And, If So, at What Cost. Cybersecurity, 2, 2
(2006).
[10] SNT. Frequently Asked Questions.
http://www.snt.utwente.nl/helpdesk/newfaq. Visited on
May 5th 2011.
[11] SURFnet. (not so) new IP features: IPv6.
http://www.ipv6.surfnet.nl/. Visited on Febuary 28th
2011.
[12] Warfield, M. H. Security Implications of IPv6. Internet
Security Systems, 2003.
[13] Yang, D., Song, X. and Guo, Q. Security on IPv6. In
Proceedings of the 2nd IEEE International Conference
on Advanced Computer Control (Shenyang, China,
March 27-29, 2010). ICACC-2010. IEEE, 2010

APPENDIX
A. QUESTIONNAIRE
The questionnaire has been added here for referencing and
possibly for future works. Note that the question routing,
which was used to make sure the respondents only saw the
questions which were applicable, was omitted.

Q1 Introduction to survey
Q2 What is your function within your company?
o Network Manager
o System Administrator
o Programmer
o Manager
o Other ____________________
Q3 What type of service or products does your company
provide? Internet services are e.g. Internet Service Providers
or services over the internet. Security Services are e.g.
firewalls etc.
o Internet Services
o Security Services
o Other ____________________
Q4 Which protocols do your services support?
￿ IPv4
￿ IPv6
Q5 Which protocols does your backbone support?
￿ IPv4
￿ IPv6
Q6 What is the reason to employ IPv6 in your services? Wat
is de reden om IPv6 in uw diensten te gebruiken?
o More secure
o Cheaper
o Anticipation of future need
o Other
Q7 What is the reason not to employ IPv6 in your services?
o Unnecessary
o Too difficult
o Too expensive
o Too unsecure
o Other
Q8 Specify the 'Other' reason to employ IPv6 in your services:
Q9 Specify the ‘Other’ reason to not employ IPv6 in your
services
Q10 When did you introduce IPv6?
o less than a year ago
o 1-2 years ago
o 2-3 years ago
o 3-5 years ago
o more than 5 years ago
Q11 What is the status of IPv6? Wat is de status van IPv6?
o Deployment ready
o Wide beta test
o Internal beta test
Q12 What IPv4 to IPv6 transition mechanism is used in your
network / services?
o Translation
o Tunneling
o Dual stack
o Other ____________________
Q13 What is the estimated amount of processed IPv6 data in
percent from total traffic processed in your services?
o 0% - 15%
o 15% - 30%
o 30% - 45%
o 45% - 60%
o 60% - 75%
o 75% - 90%
o 90% - 100%
o Unknown / n/a
Q14 What is the estimated amount of processed IPv6 data in
percent from total traffic processed in your backbone?
o 0% - 15%
o 15% - 30%
o 30% - 45%
o 45% - 60%
o 60% - 75%
o 75% - 90%
o 90% - 100%
o Unknown / n/a
Q15 Do you employ security measures for IPv4 in your
network?
o Yes
o No
Q16 Do you employ security measures for IPv6 in your
network?
o Yes
o No
Q17 How secure do you consider your IPv4 network?
o Very secure
o Secure
o Not secure nor unsecure
o Unsecure
o Very unsecure
Q18 How secure do you consider your IPv6 network?
o Very secure
o Secure
o Not secure nor unsecure
o Unsecure
o Very unsecure
Q19 What type of security measures do you employ for IPv4?
￿ Firewall
￿ Intrusion detection system
￿ Intrusion prevention system
￿ IPsec
￿ Other
Q20 What type of security measures do you employ for IPv6?
￿ Firewall
￿ Intrusion detection system
￿ Intrusion prevention system
￿ IPsec
￿ Other
Q21 What other types of security measures do you employ in
your IPv4 network?
Q22 What other types of security measures do you employ in
your IPv6 network?
Q23 What is the reason not to employ IPsec in your IPv6
network?
o Unnecessary
o Too difficult to set up
o Not much more secure
o Other ____________________
Q24 Are the rules in your security appliances consistent
between IPv4 and IPv6?
o Yes
o No
Q25 Why are the rules in your security appliances not
consistent?
o Some rules for IPv4 do not apply to IPv6
o Some rules for IPv6 do not apply to IPv4
o The network does not support IPv4
o The network does not support IPv6
o The appliances do not support IPv6
o IPv6 is considered to be less relevant / dangerous
o Too difficult to configure
o Other ____________________
Q26 Has your IPv6 network already been target of attacks?
o Yes
o No
o Unknown
Q27 What types of attacks has your IPv6 network been target
of?
Q28 Do the security appliances you use or deliver support
IPv6? Ondersteunen de beveiligingsapparaten die u gebruikt /
levert IPv6?
o Yes
o No
o Unknown
Q29 Do the security appliances you use / deliver handle IPv6
in hardware or software?
o Hardware
o Software
o Both
o Unknown
Q30 What is the reason to implement handling of IPv6 in
software?
o Cheaper to implement
o Easier to implement
o Better
o Hardware technology not available
o Other
Q31 You answered 'Other' in the question why handling IPv6
was implemented in software. Please explain what this reason
is:
Q32 How secure do you consider your IPv6 network
compared to you IPv4 network?
o Much more secure
o More secure
o Evenly secure
o Less secure
o Much less secure
Q33 Do you generally consider IPv6 to be more or less secure
than IPv4?
o IPv6 is more secure than IPv4
o IPv6 is less secure than IPv4
o They are equally safe
o No opinion
Q34 Please explain your choice:
Q35 Here you can leave any comments you wish to share: