Many people are familiar with the near-depletion of IPv4 addresses and the move to IPv6, which will free up more IP addresses for enterprises and service providers to give to new applications and devices. But the move to the new protocol opens many questions about security. Does moving to IPv6 solve the security ills of the IPv4 world, and will organizations need to throw out their existing security products once they've transitioned?
I asked quizzed security experts and co-authors of IPv6 Security, Scott Hogg and Eric Vyncke, about these and other IPv6 security issues ahead of Cisco Live in San Francisco.
Scott Hogg, CCIE, is the director of advanced technology services at GTRI and is responsible for setting the technical direction for GTRI and helping the company create service offerings in emerging technologies such as IPv6. Eric Vyncke works as a distinguished consulting engineer for Cisco Systems as a technical consultant for security covering the whole Europe. His area of expertise for 20 years is mainly security from Layer 2 to applications.
Linda Leung: Your book explains why IPv6 is already a latent threat in IPv4-only networks. That's a disturbing thought. How is IPv6 already a threat?
Scott Hogg: It has to do with the fact that many operating system manufacturers have helped aid the migration to IPv6 by including it in their OSes by default. These dual-stack operating systems perform DNS lookups, and if they get back both an IPv4 and an IPv6 address in their query they will actually prefer to communicate using IPv6 if they have the option. Most users don't realize that this capability is turned on by default and that this preference exists. Furthermore OSes like Microsoft XP, Vista, and Windows 7 will help the user establish IPv6 Internet access by dynamically creating tunnels where the IPv6 traffic will be encapsulated in IPv4 protocols to pre-determined points on the Internet. The issue is that if an organization wasn't aware of this fact, then they may actually be communicating with the Internet using IPv6 and not even realize it. These organizations probably haven't considered the security implications of this or taken any action to secure this communication in the same ways they secure their IPv4 traffic.
Eric Vyncke: Moreover, those OSes have the IPv6 protocol ready to be used: all it takes is a single IPv6 packet sent by a Layer 2 adjacent attacker pretending to be an IPv6 router. Therefore, an attacker can enable IPv6 on those OSes, and if the computer is not protected by a firewall against attacks over IPv6 and if the computer is vulnerable, then the attack will succeed.
LL: How does IPv6 improve on the security limitations of IPv4?
SH: The protocols are actually mutually exclusive, so IPv6 can't improve IPv4 security in any way. However, by learning about IPv6 security concepts you can learn about the ways you should secure IPv4. Because many of the techniques are similar (e.g. firewalls at the perimeter, access security, VPNs, and so on) the same styles of protection mechanisms we use for IPv4 can also apply to how we will protect IPv6 communications.
EV: IPv6 has also benefited from security lessons learned from IPv4. For instance, the equivalent of Address Resolution Protocol (ARP) for IPv6 is called Neighbor Discovery Protocol (NDP); by default NDP has the same security issues as ARP, but there is a secure version of NDP (called SeND) which relies on cryptography to achieve security and is available in Cisco routers.
LL: Are people lulled into a false sense of security thinking that upgrading their networks to IPv6 would be the answer to their security ills?
SH: Yes, IPv6 is a different protocol than IPv4 so it has unique challenges. IPv6 is different than IPv4 in the way that address autoconfiguration occurs. Even though IPv4 has Address Resolution Protocol (ARP) and the security weaknesses in that protocol, the weaknesses in the Neighbor Discovery Protocol (NDP) are different so they require a new method of protection. Many have the false impression that IPv6 is inherently more secure simply because it supports IPSec. IPSec is required for every IPv6 node to support but it is not mandated that is be used. IPv4 has IPSec capabilities, but they are not used for all connections.
LL: What kinds of IPv6 attacks are out there?
SH: The attacks that have gained the most popularity are in the access layer against Neighbor Discovery Protocol (NDP) and across the WAN or Internet using Routing Header 0 (RH0). There are also attacks against the transition mechanisms themselves. There is code available that anyone can download and compile to create tools that automate these attacks. That is an indication that attackers are learning about IPv6 and starting to use it as IPv6 becomes more prevalent. As IPv6 grows in usage it, will become a larger target.
For example, IPv6's weaknesses in its Neighbor Discovery Protocol (NDP) are similar to those in IPv4's Address Resolution Protocol (ARP). The security of these protocols is difficult because these protocols are essential for normal operation of the network. These attacks presume that an attacker is on our internal LAN or has already compromised a computer on the LAN. It is hard to defend against someone who already has access to your internal network. There are protocols like Secure Neighbor Discovery (SEND) and a few manufacturers like Cisco have working implementation of this protocol.
EV: The IPv6 attacks that I have seen are mainly misconfigured computers sending wrong Router Advertisements (a message normally sent by routers only) and those messages confused the network. The other attack over IPv6 that I have seen is an application-level attack (such as SSH dictionary attack or SQL injection attack or email spam) where the attacker was probably not even aware that he used IPv6 to transport the application attack.
LL: What are the security limitations of IPv6 and how should organizations protect against these vulnerabilities? Are they going to have to throw out their existing security tools for IPv6-supported gear?
SH: The IPv6 security vulnerabilities are at the access network, across the WAN and Internet. Therefore organizations must use a diverse set of techniques to secure IPv6. As an organization deployed IPv6 on the LAN they should use mechanisms to secure neighbor discovery. Tools like SEND, 802.1X, network-admission control (NAC) can be used to secure IPv6 access networks. Organizations will need to secure their tunnels with Access Control Lists (ACLs). Firewalls and IPSes can be used to secure the perimeter. These devices exist and are used today to secure IPv4, so manufacturers just need to add IPv6 capabilities to their products if they didn't already. Many of the utilities that organizations already use support both IPv4 and IPv6. It may be as simple as upgrading the software version and configuring the IPv6 protection measures. Many firewalls support both protocols, albeit IPv6 support doesn't have all the bells-and-whistles that IPv4 gets, but that gap is closing. Organizations should not need to forklift-upgrade their security devices to gain IPv6 support.
EV: Indeed, most security devices (firewall, intrusion prevention systems, VPN) exist for IPv6. The same device can even protect both IPv4 and IPv6 at the same time. It is also very important to apply the same security policy for IPv4 and IPv6 (which is easier to achieve if the same device works for IPv4 and IPv6). If the policies are different in IPv4 and IPv6, then the attackers will discover this difference and will obviously use the less strict security policy to attack your resources.
LL: During the transition process, is there any danger of security holes being exposed when organizations move from IPv4 to IPv6?
SH: Yes, during transition from IPv4 to IPv6 organizations will use a variety of transition mechanisms. While dual-stack (running both protocols simultaneously) will be the preferred approach, there will be situations that require tunneling IPv6 in IPv4 packets to help IPv6 bridge across a portion of the network that is IPv4-only. Those transition mechanisms will be the focus of attackers, and organizations will need to secure those tunnel endpoints. It depends on the organization's IPv6 transition strategy, but organizations may want to completely disable transition techniques that they are not using. The danger occurs when a transition technique like dynamic tunneling is being used without the user's knowledge and they haven't considered the implications and are not protecting the IPv6 communications.
EV:: Agreed; I have nothing to add.
LL: What should network professionals know about IPv6 security before embarking on a transition project? Are there security measures that organizations can take now even before they transition to IPv6?
SH: Network administrators should learn about the differences between IPv6 and IPv4 so they are prepared to adapt their security strategy to IPv6. Much of what they already know is applicable, but IPv6 has some nuances they should be aware of. Once they have done this they are ready to secure IPv6 ahead of when they might be deploying it. For example, an organization can disable all outbound IPv6 packets encapsulated in IPv4 packets and the dynamic tunneling techniques until such a point as they have determined if they are needed.
EV: While there are not a lot of differences between IPv4 and IPv6, it is really important that security officers learn about IPv6 now and how IPv6 can affect their security posture now (even in an IPv4-only network).
LL: The level of deployment for IPv6 is not the same across the world. For instance, it's already in use in Europe and Asia and in the U.S., ISPs are only now testing it. I'm guessing this is a kind of blessing for the U.S. since worms that usually travel across the world overnight may hit roadblocks traveling across different protocols?
SH: IPv6 is growing in popularity internationally, and most ISPs have now recognized the importance of having IPv6 services ready for their customers. Because the public and private IPv4 address space is densely populated, you can guess and IP address and it is probably being used. That is how many of today's worms operate. Due to IPv6's increased address space, it is not possible to have a worm that randomly creates target addresses as a way of spreading. You are correct in that an attack that leverages IPv4 won't affect an IPv6-only host (and vice-versa), however, most hosts will run both IPv4 and IPv6 together. Therefore, it would be possible for a worm to spread over IPv4 and then leverage IPv6 as a control channel.
EV: Also, several worms propagate over e-mail or instant messaging; and as e-mail can be sent over IPv4 or IPv6, this also means that existing worms will propagate over the IPv4 and IPv6 Internet without noticing any difference.
LL: Much of security today has to do with end-user education — teaching users not to open suspicious email attachments or fall for phishing pranks. What are the measures that end users should take to protect themselves in an IPv6 world?
SH: Since IPv4 and IPv6 are network-layer protocols, the applications that ride on top of them are the same. Therefore, if a dual-protocol application has a vulnerability in its software, that vulnerability can be exploited over IPv4 or IPv6. During the period or transition, using dual-stack organizations will bear the burden of having to protect both protocols. Unfortunately, spam, e-mail attachments, and malware-infected web sites will continue to exist in an IPv6 world. The issue is that many of the content filtering protection measures we have today only look for this malware in IPv4 packets. These products need to be looking for all these infected messages in both IPv4 and IPv6 packets.
EV: Exactly. Spam and phishing emails are already being sent over IPv6: this is completely transparent for the users and for the worm. The education is also a key component of every security policy. As Scott said, IPv6 is transparent for the end-user, so no need to educate them, but the network and security staff of enterprises and service providers must absolutely be trained about IPv6 NOW to understand the latent threats that we have just discussed.
LL: The development of Mobile IPv6 seems to have exposed more security issues than the development of IPv6. What are the security issues and how far have they been resolved in the latest standard?
SH: Mobile IPv6 represents one of the protocols that leverage the unique characteristics in IPv6 to make the system operate much better than it did using IPv4. Mobile IPv6 is being looked at to streamline mobile communications, but it must be secure before it is widely deployed. The standards specifications require that IPSec be used between the mobile node and the home agent. Some Mobile IPv6 implementations don't use IPSec to secure those control messages and that exposes a weakness. The standard has defined how Mobile IPv6 can be secured using IPSec and the return routability procedure; it is now up to implementers to use those techniques to secure Mobile IPv6 communications.
EV: Indeed, mobile IPv4 was not really the most secure and efficient protocol. Mobile IPv6 was specified to avoid all shortcomings of mobile IPv4, including security. This means that mobile IPv6 is both efficient and secure (even if not widely deployed).
LL: Let's talk about Cisco and IPv6. Cisco IOS software have been integrated with IPv6 since May 2001. Have customers been using this facility, and if so, how?
SH: When I started learning about IPv6 I would test using Cisco 2500-series routers. Back then the software worked for forwarding packets, but there weren't a lot of advanced features. Now Cisco's IPv6 implementation in IOS is very comprehensive. If you read the release notes for 12.2(33)SXI on 6500s or 12.4T on their routers, it seems that more new IPv6 features are coming than IPv4 features. Cisco has really put a lot of effort into their IPv6 implementations and it shows. Up 'til now customers have been using Cisco's IPv6 features in their test labs only. Some customers are using those IPv6 features to proceed with dual-stack deployments in parts of their networks. Customers can now feel confident that they have a set of IPv6 features in IOS that will allow them to deploy IPv6 with ease.
EV: The IPv6 stack in IOS is quite old indeed and has been tested by IPv6 users, notably universities, and in some service provider networks like NTT in Japan. Cisco is currently in the process of changing the licensing scheme for IPv6 to make it completely free; this means that if a customer buys an advanced services image for IOS, he will get exactly the same features (if applicable of course) in IPv4 or IPv6. This allows for a zero-cost deployment of IPv6 (assuming that hardware does not need to be replaced or upgraded).
LL: Most of Cisco's hardware is IPv6-enabled. How easy is it to transition to IPv6 with Cisco gear? What are the steps that customers should take?
SH: Most of Cisco's devices can forward IPv6 packets just as quickly as they can forward IPv4 packets. There isn't the performance gap between the protocols that existed five to eight years ago. It is simple to establish dual-protocol capabilities on a Cisco router or switch. You just need to assign an IPv6 address to an interface and the protocol is immediately activated. That interface will then have both an IPv4 and an IPv6 address. At that point you can enable an IPv6 routing protocol (RIPng, OSPFv3, IS-IS, EIGRPv6, or BGP) and you are ready to forward packets across your network. The next step would be to configure IPv6 DNS resource records on the DNS server to enable names to be used instead of lengthy IPv6 addresses. At that point you are ready to enable dual-protocol applications and go on your way with the migration. The migration to IPv6 doesn't require much money, but it does take a little bit of time and effort. I bet you would be surprised how easy it is if you tried it.
EV: The move to IPv6 in the network is really the easiest part, as Scott has just explained. The most difficult part is to move applications to IPv6, especially applications that directly manipulate or store IP addresses.
LL: Final question, what will you be doing at Cisco Live?
SH: I haven't been to Cisco Live in many years, so I am looking forward to attending many of the sessions. There are so many deeply technical sessions on a huge variety of technologies that my problem is deciding which sessions to attend. I will be meeting old friends and meeting new people. Eric on the other hand will be presenting on IPv6 Security like he has for several of the past years. His session seems to draw more and more people each year.
EV: Besides presenting on IPv6 Security (one version for enterprise and one version for service providers), I have also installed with some colleagues the first IPv6 network over Wi-Fi in Cisco Live. Comcast has offered us IPv6 connectivity over a 1Gbps optical fiber.
Linda Leung is an independent writer and editor in California. Reach her at firstname.lastname@example.org.