[OpenWrt-Devel] IPv6 firewall and Port Control Protocol (Was: Barrier Breaker 14.07-rc1)
david at lang.hm
Fri Jul 18 03:07:38 EDT 2014
On Thu, 17 Jul 2014, Gui Iribarren wrote:
> On 17/07/14 21:03, David Lang wrote:
>> I know that IPv6 designers pine for the "good old days" of the Internet
>> when no security was needed.
>> But the reality is that hackers and worms have shown that leaving
>> systems exposed to the Internet is just a Bad Idea.
>> As such, the idea that IPv6 would "restore" the "everyone can connect to
>> everyone on any port" of the early '80s was wishful thinking at best.
>> link-local addressing isn't a good idea, because the average home will
>> have three separate links (wired plus two bands of wireless), these can
>> get bridged together, but that causes problems as well.
>> There is no answer here that will satisfy everyone.
>> But do you really want to see the news stories about how anyone running
>> openwrt is vulnerable to $lastest_windows_exploit but people running
>> stock firmware aren't?
> Hello, thanks for joining the conversation,
> you might have not spotted this email
> as it is now, the situation is actually the opposite of what you're
> describing, it's more like: "Hey, my VoIP calls are failing, as well as
> PopcornTime videos, since I installed OpenWRT. They were working just
> fine with stock TPLink firmware"
> Have you got any examples of stock firmware that blocks incoming traffic
> by default?
> In this discussion I have only seen talk of two that don't.
Every IPv4 home router I have seen defaults to 'block all incoming, unless
something on the inside opens it'
If IPv6 routers end up being wide open, then we are going to start seeing people
getting compromized and the analysis being that it was through IPv6 and it will
get an (undeserved) reputation of being less secure than IPv4 just because
stupid vendors are going to have their stuff exposed.
We've seen worms specifically targeting printers in the past, what makes you
think we aren't going to see things like that exploiting NAS devices, DNLA
servers, thermostats, etc?
You would be horrified to see what passes for security in the Internet of
Things. A lot of that software makes me think of stuff from the '70s and early
'80s. I've seen devices manufactured in 2012 that used 4 bits for the year (with
the epoc being Jan 1 2010)!!
The horror stories that you have heard about how insecure SCADA and other
industrial devices are are not exaggerations, if anything they understate the
think about the early Internet protocols (SNMP and tFTP), and think about
systems that make them look sane and perfectly reasonable.
Exposing these systems to inbound connections from anywhere on the Internet is
Now, if these devices make a connection out to phone home, allowing that home to
reach back is reasonable, and supporting things like upnp to allow devices to
specifically open up inbound connections are reasonable. I'm not saying that it
needs to be as hard to configure as getting in through IPv4 NAT, but it should
NOT be the 'open end-to-end Internet the way $DIETY intended'
look at how easy it is to 'root' phones, where the company involved is at least
reasonable competent in writing software. For a lot of the IoT devices, the
Internet is a rushed, tacked on addition (they already needed a processor to
manage something, so spend a few cents more and now they can advertise this
mobile device app). Try using some of these apps and devices and see how
horrific the software is.
>> Yes, it would be ideal if every host was locked down so that it was safe
>> for them to be exposed.
>> But that's not the world we live in.
>> David Lang
>> On Wed, 16 Jul 2014, Lyme Marionette wrote:
>>> ----- Original Message -----
>>> On Wednesday, July 16, 2014 2:10:53 PM "Gui Iribarren"
>>> <gui at altermundi.net> wrote:
>>>> Benjamin is giving some great examples of real-world scenarios where
>>>> default-open firewall simplifies administration,
>>>> and where a default-closed firewall would be not only unnecessary
>>>> (provides no benefits), but would indeed complicate setting up
>>> There have been many good arguments posted on this subject and to
>>> throw my opinion in, it a question of effort and expectations.
>>> I think everyone can agree that:
>>> -It takes equal effort to turn a firewall on, as it does to turn one off.
>>> -It takes equal effort to create a specific block list, as it does to
>>> create a specific allow list.
>>> -UPnP is not included by default for either the ipv4 or ipv6 stacks.
>>> I would also go further to suggest that:
>>> -Consistency is good, even if it consistent for superficial reasons.
>>> We know that, for NAT reasons, that the ipv4 stack by default blocks
>>> incoming connections:
>>> -Because it doesn't know by default where to route them.
>>> -ipv4 end-points have been traditionally insecure.
>>> The two ways to get around this (for gaming, etc):
>>> -Through setting firewall rules to route the traffic to an end-point.
>>> -Through the use of UPnP (which is used by most games to host, and
>>> gaming consoles).
>>> With the adoption of ipv6 there is the opportunity to change this
>>> behaviour such that instead of incoming traffic being restricted for
>>> technical reasons, that incoming traffic is routed to the correct
>>> However, that begs the questions:
>>> A) Is that consistent with what people would expect?
>>> B) Are ipv6 end-points secure by design?
>>> In regards to A, from the mindset of a non-technical user, would wager
>>> that the answer is 'no'. Even though there is a change in technology
>>> with ipv6, the ipv6 technology fulfills the same role as ipv4 and this
>>> could be seen as opposing direction between the two. This would likely
>>> catch many end-users by surprize unless they read the small print
>>> regarding this.
>>> As for B, given my view of software development, applications,
>>> networks, etc (I've been in the IT business for over 25 years now) I
>>> would wager that 80% of applications are secure, and that the 0ther
>>> 20% make the potential change in policy very risky.
>>> IMO, which others may disagree with, would be to include UPnP by
>>> default which would/should resolve most of the hosting issues.
>>> openwrt-devel mailing list
>>> openwrt-devel at lists.openwrt.org
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
More information about the openwrt-devel