In the Usenet newsgroup comp.os.linux.networking, in article
<(E-Mail Removed)>, T. Little wrote:
>We're doing 3DES between the ATMs and the controller.
Ah, OK - you have clue. You might not believe some of the requests that
get posted to Usenet.
>The ATMs are behind cable modems, but the cable modems do not pass
>traffic to the Internet - - the traffic goes to our cable provider's NOC
>and gets VLAN'd to our DP center. The cable provider calls this "TLS,"
>and it's fast and cheap compared to using, say, individual 56k frames or
ISDN..
That's 802.1Q - something similar to protocol 97 (0x61), a.k.a.
Ethernet-within-IP Encapsulation RFC3378.
3378 EtherIP: Tunneling Ethernet Frames in IP Datagrams. R. Housley,
S. Hollenbeck. September 2002. (Format: TXT=18803 bytes) (Status:
INFORMATIONAL)
>But my concern has to do with someone figuring out our topology and
>getting busy with their own laptop.
They're not going to be able to impersonate the ATM, because they won't
have the 3DES key - so that's out. Being able to send datagrams to your
system? To what end? Denial of Service? One presumes that the only
connection you'd accept from (call it) the ATM network is ATM stuff
which would be protected by the 3DES. Or do the ATM machines like to
surf pr0n in between customers ;-)
>I figure that anyone getting access to our cable will have to move fast
>- - we'll be paged as soon as a connection is physically broken, and we
>have cameras on both sides (public and service sides) of the ATM units.
>I'd just like to buy us as much time as possible for that day when someone
>tries to break in.
I see where you are coming from - but I don't think this is the right
mechanism. Actually, what might even be simpler for you is arpwatch
[compton ~]$ apropos arpwatch
arpsnmp [arpwatch] (8) - keep track of ethernet/ip address pairings
[compton ~]$
You don't mention a Linux distribution other than IPCOP, but arpwatch is
often included in regular Linux distributions.
181922 May 20 18:25 arpwatch-2.1a13-12.i386.rpm
That's Fedora Core 4. We do something very similar, monitoring the ARP
caches of routers, servers and the network switches. Company policy does
not permit visiting computers. Should one get in, our adaptation of this
program causes pop-up messages on select computers in the NOC and Security.
This also includes (thanks to the network switches) an indication of which
network drop is involved. This brings the Thundering Herd Of People Who Do
Not Smile - often before the intruder can finish booting.
>The ATMs themselves are encased in concrete and steel, but I don't even
>trust the Diebold and Brinks guys who service the ATM units.
I dunno about Brinks (they're Brinks problem - and the bonding company),
but I don't know that many people overly thrilled by Diebold. In your
Copious Free Time(tm), do a search for the word 'Diebold' in the Usenet
news group 'comp.risks' which is a mirror of the "ACM FORUM ON RISKS TO
THE PUBLIC IN COMPUTERS AND RELATED SYSTEMS".
>I know that the ATM service guys have the IP and MAC information for
>our units already. I guess the question is what we can do to thwart
>someone who has both physical access and fundamental network
>information.
I'm not sure you can do that much using IP or MAC addresses. Both can be
spoofed with exceptional ease. A _MINOR_ possibility would be to quietly
tweak the network parameters in the TCP stack, (example, changing TTL, or
certain flags) and then monitoring this with a passive O/S fingerprinting
tool like p0f (
http://lcamtuf.coredump.cx/p0f.shtml), but the additional
security would likely be minimal. It's more a "Security by Obscurity"
feature than anything else.
>Unfortunately, the controller and ATMs - - amazingly - - are Windows based.
Well aware of it - again, see that 'comp.risks' newsgroup.
>They used to be OS/2 talking via serial lines to an RS-6000 running AIX.
>And it was rock solid, I might add.
Yeah, I switched banks when my former bank "upgraded".
>I'll have a look at your static ARP suggestion, though I'm not sure
>about implementation with Windows hosts.
As above - it may not be relevant any more.
>Tomorrow, I plan to play with IPCOP a bit. The MAC criteria is seeming
>not so important as I'd thought.
IPCOP is a user friendly front end for configuring the netfilter code in
a Linux kernel. As such, it does things that the application author
thought would be useful, and doesn't do what he didn't think about, or
didn't include in his interface.
Old guy