[plug] Safely using an untrusted router

Dean Bergin dean.bergin at gmail.com
Thu Oct 22 23:46:17 UTC 2015 

Hello Dirk,

>From a network perspective, as long as the modem is in bridged mode (ie.
pass-through - think of it as a dumb 2 port switch). It is only interested
in layer 2 and above for the purposes of learning mac address and knowing
which interface frames need to be sent out of. The device can still pass
layer3 and above, but from the bridge perspective, it only really cares
about frames on either ADSL/LAN port(s) and passes any higher layers on
through without inspection or consideration.

Not to mention that the device itself can actually be configured to have a
completely different LAN IP settings to that of your own LAN (effectively
isolating it from your network to layer2 frames only) and therefore it
should never be able to even do anything useful from a networking
perspective especially since there would be no routes from the device to
any other network whatsoever, therefore it would not be able to participate
in any botnet, phone home or any such activity (including being able to
communicate on port 32764 as this would require proper layer3
configuration, which we simply deny in this case), unless any issues you
may be experiencing are being caused by it's failure to be able communicate
over layer 3 on the device, or there is a much, much, much more
sophisticated hack in play here (still unlikely since SOHO equipment is
very cheap and often have very low resources, such as RAM/CPU).

If there is any funny business happening on the device, I would suspect
that the device is likely faulty. Networked device nodes with faulty
network controllers have a habit of messing up layer2 frames, but in the
case described above, it should mitigate any perceived network threat with
the device to a very large degree, depending of course how any possible
hacks have been implemented.

I doubt hackers would be interested in messing with layer2/framing though
as this would probably break the bridging functionality.

You could still look into a VPN solution, but if the device is messing with
any traffic, then you could find yourself running into all sorts of issues
with trying to establish and maintain a VPN connection anyway.

I would also suggest running wireshark (with the assistance of another IT
professional, if need be) to do some packet sniffing and see if the device
is indeed misbehaving. To me, it seems more logical (read: less effort) to
at least test with a different modem.


On Thu, Oct 22, 2015 at 8:57 PM Dirk <justanothergreenguy at gmail.com> wrote:

> Hi Bill,
>
> Your setup probably makes a lot of sense for your needs, but I don't need
> to remote out or remote in, or support VoIP, or support multiple
> platforms.  It's just one PC and a modem router.  There should be a simple
> solution.
>
> I wipe and restore my throwaway Linux OS and swap part'n after each
> online session, and restore a fresh trusted MBR each time.  The
> throwaway OS is updated offline using verified package lists and security
> update .deb's from previous sessions (from CD, not USB).  My Mint ISO is
> a verified good copy (MD5).  And yet I'm experiencing corrupted package
> lists, so I can only suspect my BIOS or the router.  I don't think BadUSB
> exploits are circulating as yet, so the mouse and keyboard should be fine.
> I'm careful where I browse and what I click on when using this PC.  Java,
> Flash, JavaScript (unless req'd), application launching in the
> browser, linked fonts, smbd, avahi-daemon, auto-mounting, etc etc are all
> disabled.  I'm far less concerned when I'm using other devices with my wifi
> router, as I don't log into anything important with them.
>
> So my consumer / SOHO modem router is by far the weakest link.  It's
> secured as best as possible (re UPnP and WAN-side admin disabled, strong
> admin password, no wireless functionality, unsolicited packets are
> dropped, etc etc), but the firmware is well out of date (no surprises
> there, most are), and hackers are turning their attention to SOHO routers,
> coz most people wouldn't know it if their router was hacked and added to a
> botnet (or whatever);  e.g. AV doesn't reach into the router.  SOHO routers
> are the new low-hanging fruit for hackers.
>
> So I don't think it's all that unreasonable for a SOHO to regard their
> SOHO router as untrusted (i.e. like anyone needing to connect thru an
> untrusted hotel (W)LAN, for instance), and to seek a solution that allows
> for such untrusted elements, as per my original post.
>
> You are of course right.  There are many many areas of risk apart from the
> router.  But what other vulnerabilities can I address?  What do you mean by
> 'real risks', if not what I'm already addressing above?
>
>
>
>
>
> On Thursday, 22 October 2015, BillK <billk at iinet.net.au> wrote:
>
>> The VPN is set up for multiple clients in routed mode. I regularly use an
>> android phone and tablet, occasionally a windows desktop and site to site
>> links. I did set up an iPad at one time. Carries ssh, email, calendaring,
>> sip VoIP, security video etc all on private networks. It's all tied
>> together by ospf on the router and various gentoo Linux hosts including the
>> VPN concentrator VM.
>>
>> I also use a stunnel instance in the VM with proxy tunnel and putty on
>> windows to tunnel ssh out of heavily locked down networks. Both openvpn and
>> tunnel listen on public non-standard ports port forwarded through the
>> router. SSL is also port forwarded to an ssl multiplexor in the VM
>> listening on port443 to redirect incoming SSL to either openvpn or stunnel
>> as required, the end points terminate as SSH on my desktop.
>>
>> It sounds complex, isn't really and is quite useful.
>>
>> However what has bothered me about this thread is the emphasis on the
>> router as a problem ... It generally isn't, a router is not automatically
>> compromised so fix that vulnerability first, then attend to real risks.
>> These days you are at far more risk from perverted/subverted mobile devices
>> ... In the scheme of things routers are just one of, and definitely not the
>> main thing you have to worry about.
>>
>> BillK
>>
>>
>> On 22 October 2015 4:25:51 pm AWST, Dirk <justanothergreenguy at gmail.com>
>> wrote:
>>>
>>> Thanks Bill,
>>>
>>> Your setup looks a bit too complicated for me as well :)  ...but good to
>>> know iiNet are assigning private IP addresses to their mobile users.
>>>
>>> Is the VPN just between your VM and your phone?  Interesting
>>> idea, although I'm not sure Android would be the safest bet :)  ...oops, I
>>> hope I didn't start any Android v iPhone pie-slinging :)
>>>
>>> Anyway, I think I'm going to pursue the 4G USB modem idea for now, and
>>> see how I go.
>>>
>>> Thanks everyone so much for your ideas and comments.  I didn't mean
>>> for my wee little Qu to dominate the PLUG forum, and I kinda feel like
>>> I'm stretching my welcome a little bit for a first time contributor, so I
>>> do apologise to all who put up with me the last few days.
>>>
>>> But if the group is ok with it, and anyone has any further ideas, or
>>> feedback on the 4G USB modem approach, please don't hesitate to let me know.
>>>
>>> Cheers, Dirk
>>>
>>>
>>>
>>>
>>> On Thursday, 22 October 2015, Bill Kenworthy <billk at iinet.net.au> wrote:
>>>
>>>> my iinet mobile (GalaxyS5 with cyanogenmod) is using a 10.160 private
>>>> address - I have no problems running openvpn over it to a vm inside my
>>>> network (including security camera video over the vpn when I want to
>>>> take a peek).  I am using a Cisco 1841 with an alcatel speadstream
>>>> bridged with the router doing pppoe to iinet.  Stable but I need to work
>>>> on the QoS more.
>>>>
>>>> BillK
>>>>
>>>>
>>>>
>>>> On 22/10/15 07:12, Dean Bergin wrote:
>>>> > Hello Dirk,
>>>> >
>>>> > This is probably not going to help solve your particular issue, but
>>>> one
>>>> > thing I recently did, was install OpenWRT on a Rpi2 and set up PPPoE
>>>> > over one of two subinterfaces (VLAN) to a cheap netgear modem (with
>>>> the
>>>> > help of a Cisco Catalyst switch). I also put the Rpi2 OpenWRT
>>>> > effectively into it's own routed subnet/DMZ (part of the design) so
>>>> that
>>>> > even if there where to be some kind of funny business, things like
>>>> uPNP
>>>> > theoretically should not work since my experience has taught me that
>>>> > most consumer-grade modems/routers do not route/NAT anything other
>>>> than
>>>> > their resident subnet, therefore I believe that not only are uPNP
>>>> > implementations (and many other services on consumer-grade routers)
>>>> > usually bound to the subnet to which they are running on, but should
>>>> be
>>>> > disabled in cases where the device is in pass-through mode.
>>>> >
>>>> >>Does anyone know whether 4G modems (and smart phones, for that matter)
>>>> > are assigned a publicly-routable IP address or are they
>>>> > typically NAT'd behind a small number of IP addresses of the mobile
>>>> > service provider's servers?  I can't imagine billions(?) of mobile
>>>> > phones all having unique publicly-routable IP addresses (on top of all
>>>> > the servers and so on, around the world).
>>>> >
>>>> > I had the opportunity to test this, as I was able to tether my phone
>>>> to
>>>> > a Rpi2 running OpenWRT as part of the labs I did for my now current
>>>> > nework design, but I did not think to test this specific scenario.
>>>> >
>>>> > Shouldn't be too difficult to create a lab to test this, if someone
>>>> has
>>>> > a spare raspberry pi (mine is currently in 'prod' now)?
>>>> >
>>>> >
>>>> > On Wed, Oct 21, 2015 at 6:27 PM Dirk <justanothergreenguy at gmail.com
>>>> > <mailto:justanothergreenguy at gmail.com>> wrote:
>>>> >
>>>> >     Thanks Andrew.  Will follow up on those ideas too, thanks.
>>>> >
>>>> >     However, I have another idea, a bit left field, but it may just do
>>>> >     the trick...
>>>> >
>>>> >     Does anyone know whether 4G modems (and smart phones, for that
>>>> >     matter) are assigned a publicly-routable IP address or are they
>>>> >     typically NAT'd behind a small number of IP addresses of the
>>>> mobile
>>>> >     service provider's servers?  I can't imagine billions(?) of mobile
>>>> >     phones all having unique publicly-routable IP addresses (on top of
>>>> >     all the servers and so on, around the world).
>>>> >
>>>> >     If they're NAT'd, then maybe a pre-paid 4G USB modem dongle would
>>>> be
>>>> >     the way to go for low MB critical online work, eg. fetching
>>>> package
>>>> >     lists, logging in to ASIC, ATO, webmail, our utilities, etc.
>>>> Should
>>>> >     block all scanners on the net that are looking for routers to
>>>> >     exploit, by virtue of sitting behind the Svc providers routers.
>>>> >      (...and then use an unsecured computer and ADSL router pair for
>>>> >     general web browsing, content streaming, etc).
>>>> >
>>>> >     Does anyone know if this would work?
>>>> >
>>>> >     (Of course, if a 4G dongle is not NAT'd then I don't really gain
>>>> >     anything).
>>>> >
>>>> >
>>>> >
>>>> >
>>>> >     On Wednesday, 21 October 2015, Andrew Cooks <acooks at gmail.com
>>>> >     <mailto:acooks at gmail.com>> wrote:
>>>> >
>>>> >         On Wed, Oct 21, 2015 at 9:43 AM, Dirk
>>>> >         <justanothergreenguy at gmail.com> wrote:
>>>> >
>>>> >
>>>> >             Cheers for that Pavel.  And thanks again Brad for your
>>>> >             input.  You've both given me some ideas, although I was
>>>> >             hoping for an easy OpenVPN option  :)
>>>> >
>>>> >             If anyone else has any thoughts or suggestions, please let
>>>> >             me know!
>>>> >
>>>> >         My internet access is slow enough, so I'm not really excited
>>>> >         about pushing everything through a VPN.
>>>> >
>>>> >         I trust my router. I have a TP-Link TD-8817 modem in bridge
>>>> >         mode, connected to a fit-pc
>>>> >         (http://www.fit-pc.com/web/solutions/multilan/) running
>>>> IPFire
>>>> >         (http://www.ipfire.org/). IPFire tells me I can trust my DNS.
>>>> >         IPFire packages are kept up to date. The modem could
>>>> conceivably
>>>> >         modify the PPPoE frames in transit, except that it's a dirt
>>>> >         cheap consumer product with little functionality that could be
>>>> >         exploitable and it's unlikely to have enough processing power
>>>> to
>>>> >         do that kind of thing.
>>>> >
>>>> >         There is nowhere safe, only acceptable risks.
>>>> >
>>>> >         a.
>>>> >
>>>> >     _______________________________________________
>>>> >     PLUG discussion list: plug at plug.org.au <mailto:plug at plug.org.au>
>>>> >     http://lists.plug.org.au/mailman/listinfo/plug
>>>> >     Committee e-mail: committee at plug.org.au <mailto:
>>>> committee at plug.org.au>
>>>> >     PLUG Membership: http://www.plug.org.au/membership
>>>> >
>>>> > --
>>>> >
>>>> > Kind Regards,
>>>> >
>>>> > /Dean Bergin/.
>>>> >
>>>> >
>>>> >
>>>> > _______________________________________________
>>>> > PLUG discussion list: plug at plug.org.au
>>>> > http://lists.plug.org.au/mailman/listinfo/plug
>>>> > Committee e-mail: committee at plug.org.au
>>>> > PLUG Membership: http://www.plug.org.au/membership
>>>> >
>>>>
>>>> _______________________________________________
>>>> PLUG discussion list: plug at plug.org.au
>>>> http://lists.plug.org.au/mailman/listinfo/plug
>>>> Committee e-mail: committee at plug.org.au
>>>> PLUG Membership: http://www.plug.org.au/membership
>>>>
>>> ------------------------------
>>>
>>> PLUG discussion list: plug at plug.org.au
>>> http://lists.plug.org.au/mailman/listinfo/plug
>>> Committee e-mail: committee at plug.org.au
>>> PLUG Membership: http://www.plug.org.au/membership
>>>
>>>
>> --
>> Sent from my Android phone with K-9 Mail. Please excuse my brevity.
>>
> _______________________________________________
> PLUG discussion list: plug at plug.org.au
> http://lists.plug.org.au/mailman/listinfo/plug
> Committee e-mail: committee at plug.org.au
> PLUG Membership: http://www.plug.org.au/membership

-- 

Kind Regards,

*Dean Bergin*.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.plug.org.au/pipermail/plug/attachments/20151022/09a4abef/attachment.html>

More information about the plug mailing list