IPv4 Hijacking For Idiots

Mel Beckman mel at beckman.org
Mon Jun 5 16:28:11 CST 2017


I didn’t research Ron’s specific example. I was speaking in generalities. I’m assuming any BGP hijacker already has two or more DIA connections. It only costs $100 to add BGP peering to that setup. Yes, they will need an ASN. I was only talking about the cost of adding an upstream BGP session.


On Jun 5, 2017, at 9:03 AM, Christopher Morrow <morrowc.lists at gmail.com<mailto:morrowc.lists at gmail.com>> wrote:

On Mon, Jun 5, 2017 at 7:05 AM, Mel Beckman <mel at beckman.org<mailto:mel at beckman.org>> wrote:
One way is for the hijacker to simply peer with himself. The hijacker has an existing peering arrangement with, say, AT&T. He then tells AT&T that he will be transit for ASxxxx advertising XYZ routes, by dint of a cheerfully forged LOA. Once filters have been updated, the hijacker advertises the space to himself, and then from thence to AT&T.

that doesn't seem to be what's happening in ron's example though...

it looks, to me, like the example ron has is more a case of:
  1) register contacts for lost asn (AS34991)
  2) setup equipment/etc at an IX (bulgaria-ix it seems, at least) with another shill/lost-child asn (AS206776)
  3) start doing the bgps with the IX fabric's route-server
  4) profit (or something)

so here the IXP operator (balkans ix actually?)
  (search for 206776 -> http://lg.bix.bg/?query=bgp&addr=neighbors+

should probably look more than just side-eyes at their customer...

It's no great trick getting peering set up. Just fill out a ten-question BGP app and pay a one-time fee of maybe $100, and you're done.

err, you'll have to better explain this I think.

Are you saying: "get an ASN from RIR that costs 100USD" (might, probably does)

this doesn't get you a peering/transit contract though...


 -mel beckman

> On Jun 5, 2017, at 3:56 AM, Ronald F. Guilmette <rfg at tristatelogic.com<mailto:rfg at tristatelogic.com>> wrote:
> The more I know, the less I understand.
> Maybe some of you kind folks can help.
> Please explain for me the following scenario, and how this all actually
> works in practice.
> Let's say that you're a malevolent Bad Actor and all you want to do is
> to get hold of some ASN that nobody is watching too closely, and then
> use that to announce some routes to some IPv4 space that nobody is
> watching too closely, so that you can then parcel out that IP space
> to your snowshoe spammer pals... at least until somebody gets wise.
> OK, so you pull down a copy of, say, the RIPE WHOIS database, and you
> programatically walk your way through it, looking for contact email
> addresses on ASN records where the domain of the contact email address
> has become unregistered.  Say for example the one for AS34991.  So
> then you re-register that contact domain, fresh, and then you start
> telling all of your friends and enemies that you -are- AS34991.
> That part seems simple enough, and indeed, I've seen -this- part of the
> movie several times before.  However once you have stepped into the
> identity of the former owners of the ASN, if you then want to actually
> proceed to -announce- some routes, and actually ave those routes make
> it out onto the Internet generally, then you still have to -peer- with
> somebody, right?
> So, I guess then, if you're clever, you look and see who the ASN you've
> just successfully hijacked has historically peered with, and then you
> somehow arrange to send route announcements to those guys, right?
> (I'm talking about AS206776 and    AS57344 here, BTW.)
> But see, this is where I get lost.  I mean how do you push your route
> announcements to these guys?  (I don't actually know that much about
> how BGP actually works in practice, so please bear with me.)  How do
> you know what IP address to send your announcements to?  And if you are
> going to push your route announcements out to, say, the specific routers
> that are run by AS206776 and AS57344, i.e. the ones that will send your
> desired route announcements out to the rest of the Internet... well..
> how do you find out the IP addresses of those routers on those other
> networks?  Do you call up the NOCs at those other networks and do a bit
> of social engineering on them to find out the IP addresses you need to
> send to?  And can you just send BGP messages to the routers on those
> other networks without -any- authentication or anything and have those
> routers just blindly accept them -and- relay them on to the whole rest
> of the Internet??
> I've read article after article after article bemoanging the fact that
> "BGP isn't secure", but now I'm starting to wonder just how massively
> and unbelieveably unsecure it actually is.  I mean would these routers
> being run by AS206776 and AS57344 just blindly accept -any- route
> announcements sent to them from literally -any- IP address?  (That seems
> positively looney tunes to me!  I mean things can't really be THAT
> colossally and unbelievably stupid, can they?)
> Thanks in advance for any enlightenment.
> Regards,
> rfg
> P.S.  It would appear to be the case that since some time in April of this
> year the "Bulgarian" network, AS34991, had evinced a rather sudden and
> pronounced affinity for various portion of the IPv4 address space nominally
> associated with the nation of Columbia, including at least five /24 blocks
> within<> which, from where I am sitting, would appear to belong
> to the National University of Columbia.
> Oh well.  They apparently haven't been missing those five gaping holes in
> their /16 since the time the more specifics started showing up in April.
> And anyway, so far it looks like the new owners of AS34991 haven't actually
> sub-leased any of those /24s to any spammers yet.  Only the<>
> block seems to be filled, wall-to-all, with snowshoe spammers so far.

More information about the NANOG mailing list