IPv6 Focus Month: Device Defaults - SANS Internet Storm Center

IPv6 Focus Month: Device Defaults
IPv6 in this part of the planet is not very advanced, as in the deployment isn't. Whilst companies and telcos realise that the end so to speak is nigh for IPv4 uptake is rather slow in AU at least. Telcos are however quickly addressing this and no doubt a number of them are close to enabling IPv6 to your gateway. If they haven't already. This brings be to my favourite devices, firewalls. During a bunch of security reviews over the last year or so we typically spend a little bit of time looking at the IPv6 setups and requirements in the organisations. We certainly found that people quite readily state they have no IPv6 in their environment, however often when they RDP, SSH or otherwise connect to a more recent version of insert your favourite OS here, the connection is most definately IPv6. When you then look at firewall configurations you often find nice looking IPv4 rules to control traffic and a less than ideal default for IPv6 ANY, ANY, ANY permit. So does that mean when your telco enables IPv6 to your gateway, traffic can leave? Potentially yes, it does depend on a number of other factors, but the core of it is that people do not realise they may be leaking. Even if traffic to the internet is restricted, what about other network segments? In a PCI DSS pentest, connectivity via IPv4, nope, nicely segmented. IPv6 please come through, full access. Another thing to remember with firewalls is that IPv6 is relatively new to them as well. So maybe you need to check out whether your product does support IPv6 and if the answer is yes, to what extent. What about other devices in the network, your switches and routers. will their current or even latest OS support you IPv6 requirements. Printers, Multifunction devices etc, do they support it. Do they have defaults that really do not help you out from a security perspective. For today that is what I would like to hear from you. What devices have you come across that have "interesting" IPv6 defaults. Maybe they don't support it fully. Maybe they just get it wrong. One firewall a few years ago (fixed now) did IPv6 to IPv4 translation a bit diferrently and mangled the IPv4 packets that resulted. So what are your IPv6 watch out for this tips? Mark	Mark 392 Posts ISC Handler Mar 5th 2013
Thread locked Subscribe	Mar 5th 2013 9 years ago
Thinking out loud... I have an iptables f/w at home masquerading an RFC1918 network behind a single IP. What happens when my carrier brings IPv6 to the gateway? Will the same setup work? Does NAT/Masquerading work with IPv6? Or will I get a v6 subnet and have to let all my internal devices route to the carrier's gateway? (It has a firewall of sorts -- today w/ v4 -- but I don't tend to rely on it.) Would a transparent (bridging) firwall running ip6tables work between me and the gateway? I use things like DNAT on my firewall to transparently snag outbound HTTP and DNS traffic to route them to my own HTTP proxy and internal resolver (forcing the kids to DansGuardian and a resolver that uses OpenDNS for filtering). Will that work in an IPv6 universe? That's all related to my home environment, but I'm sure the same issues will exist for many enterprises.	Hal 50 Posts
Quote	Mar 5th 2013 9 years ago
@HAL: connection tracking forLinux came with 2.6.20 Don't know whether this was backported to the maybe older kernel within your distro,	Jens 42 Posts
Quote	Mar 6th 2013 9 years ago
This is not a built-in vulnerability as much as a configuration issue. The Microsoft ISA firewall, at least the early versions, was solidly 100% IPv4. If you enabled IPv6 on the underlying Windows operating system, it happily passed IPv6 traffic among all the interfaces and nary a thing showed up in the ISA logs.	Anonymous
Quote	Mar 6th 2013 9 years ago
Although it's out of support, Red Hat Enterprise Server 4 and probably CentOS v4 enabled IPv6 by default when installed. There was no indication in the installer at all. It just worked.	Anonymous
Quote	Mar 6th 2013 9 years ago
@HAL: Forget NAT, every host in your home network gets one (or more) fully routable address(es) plus gateway info. Thus indeed: the firewall on the router is more important now. Don't forget: All Android and iOS devices do v6 by default as well. Even my HP Printer does! Btw: Subnet in v6 is called a suffix and usually means you'll end up with a /48 or /56 slice. And that's just in your home network	Anonymous
Quote	Mar 10th 2013 9 years ago
@Mark: Regarding devices with interesting v6 defaults: Apple iPad 2 uses ipv6 but appears to send DNS requests using ipv4 only. It also appears to favour v4 over v6 but that could come from the happy eyeballs implementation in safari. Which in itself is probably caused by the fact that some sites show higher return trip times in v6 than in v4 (facebook for instance).	Anonymous
Quote	Mar 10th 2013 9 years ago

IPv6 in this part of the planet is not very advanced, as in the deployment isn't. Whilst companies and telcos realise that the end so to speak is nigh for IPv4 uptake is rather slow in AU at least. Telcos are however quickly addressing this and no doubt a number of them are close to enabling IPv6 to your gateway. If they haven't already. This brings be to my favourite devices, firewalls.

During a bunch of security reviews over the last year or so we typically spend a little bit of time looking at the IPv6 setups and requirements in the organisations. We certainly found that people quite readily state they have no IPv6 in their environment, however often when they RDP, SSH or otherwise connect to a more recent version of insert your favourite OS here, the connection is most definately IPv6. When you then look at firewall configurations you often find nice looking IPv4 rules to control traffic and a less than ideal default for IPv6 ANY, ANY, ANY permit. So does that mean when your telco enables IPv6 to your gateway, traffic can leave? Potentially yes, it does depend on a number of other factors, but the core of it is that people do not realise they may be leaking. Even if traffic to the internet is restricted, what about other network segments? In a PCI DSS pentest, connectivity via IPv4, nope, nicely segmented. IPv6 please come through, full access.

Another thing to remember with firewalls is that IPv6 is relatively new to them as well. So maybe you need to check out whether your product does support IPv6 and if the answer is yes, to what extent.

What about other devices in the network, your switches and routers. will their current or even latest OS support you IPv6 requirements. Printers, Multifunction devices etc, do they support it. Do they have defaults that really do not help you out from a security perspective.

For today that is what I would like to hear from you. What devices have you come across that have "interesting" IPv6 defaults. Maybe they don't support it fully. Maybe they just get it wrong. One firewall a few years ago (fixed now) did IPv6 to IPv4 translation a bit diferrently and mangled the IPv4 packets that resulted. So what are your IPv6 watch out for this tips?

Mark

Mark

392 Posts
ISC Handler

Mar 5th 2013

Thinking out loud... I have an iptables f/w at home masquerading an RFC1918 network behind a single IP. What happens when my carrier brings IPv6 to the gateway? Will the same setup work? Does NAT/Masquerading work with IPv6?

Or will I get a v6 subnet and have to let all my internal devices route to the carrier's gateway? (It has a firewall of sorts -- today w/ v4 -- but I don't tend to rely on it.) Would a transparent (bridging) firwall running ip6tables work between me and the gateway?

I use things like DNAT on my firewall to transparently snag outbound HTTP and DNS traffic to route them to my own HTTP proxy and internal resolver (forcing the kids to DansGuardian and a resolver that uses OpenDNS for filtering). Will that work in an IPv6 universe?

That's all related to my home environment, but I'm sure the same issues will exist for many enterprises.

Hal

50 Posts

@HAL: connection tracking forLinux came with 2.6.20
Don't know whether this was backported to the maybe older kernel within your distro,

Jens

42 Posts

This is not a built-in vulnerability as much as a configuration issue. The Microsoft ISA firewall, at least the early versions, was solidly 100% IPv4. If you enabled IPv6 on the underlying Windows operating system, it happily passed IPv6 traffic among all the interfaces and nary a thing showed up in the ISA logs.

Anonymous

Although it's out of support, Red Hat Enterprise Server 4 and probably CentOS v4 enabled IPv6 by default when installed. There was no indication in the installer at all. It just worked. :-)

Anonymous

@HAL: Forget NAT, every host in your home network gets one (or more) fully routable address(es) plus gateway info. Thus indeed: the firewall on the router is more important now.
Don't forget: All Android and iOS devices do v6 by default as well. Even my HP Printer does!
Btw: Subnet in v6 is called a suffix and usually means you'll end up with a /48 or /56 slice.
And that's just in your home network ;-)

Anonymous

@Mark: Regarding devices with interesting v6 defaults: Apple iPad 2 uses ipv6 but appears to send DNS requests using ipv4 only. It also appears to favour v4 over v6 but that could come from the happy eyeballs implementation in safari.
Which in itself is probably caused by the fact that some sites show higher return trip times in v6 than in v4 (facebook for instance).

Anonymous