so we started to wonder if, since we started protecting our bgp
sessions with md5 (in the 1990s), are there still folk trying to
attack?
we were unable to find bgp mib counters. there are igp interface
counters, but that was not our immediate interest. we did find
that md5 failures are logged.
looking at my logs for a few years, i find essentially nothing;
two 'attackers,' one my own ibgp peer, and one that noted evildoer
rob thomas, bgprs01.ord08.cymru.com.
we would be interested in data from others.
note that we are neither contemplating nor suggesting removing md5
from [y]our bgp sessions.
randy
Just to point out -
Data about md5 attacks from various organizations will depend on a number of factors such as -
Is BGP TTL Security check being done?
Are anti-spoofing ACLs enabled?
uRPF enabled? Strict or Loose?
BGP Session over a separate interface (tunnel)?
With Gratitude,
Pratik Lotia | Security Engineer | Advanced Engineering Security
Charter Communications
"A satisfied customer is the best business strategy of all."
n00b response here
I thought using ACLs or otherwise protecting the BGP endpoint was best practice. Thus it's really hard to even try break an MD5 protected BGP session if you can't even establish the TCP connection.
Everything that I've seen or set up had an ACL to only allow the peer(s) to be able to connect to (from memory) TCP port 179.
Is there something that I've missed the boat on?
#learningOpportunity
Well, think about RST attacks, in which someone bombards a TCP connection with TCP RESET in the hopes of threading a needle and taking it down. It's not the end of the world - BGP restarts - but there is an outage. The simplest way to protect against that (and against having someone with a hijacked IP address connect to your router) is to put mutual authentication on the TCP connection. Having it also at the BGP layer, and having ACLs to be sure you know what's going on, are good things, but TCP MD5, TCP-AO, or IPsec are an awful lot safer.
To further harden your setup, consider using GTSM
https://tools.ietf.org/html/rfc5082
Kind regards,
Job
No - it's a belt-and-suspenders sort of thing, along with GTSM.
so we started to wonder if, since we started protecting our bgp
sessions with md5 (in the 1990s), are there still folk trying to
attack?
To recap for the purpose of my own edification and because hopefully
someone will relieve me of my assumptions.
The purpose of of rfc 2385 tcp md5 digests is to keep in-window, tcp
segments that are spoofed from being ingested into the tcp stack. At the
time of it's writing (1998) some popular network operating systems did
not check that the sequence number was in fact inside the window (so
that any tcp packet matching the 4 tupple would be ingested whether it
was in-window or not. Variously this improvement was supplemented with
the checking the TTL (https://tools.ietf.org/html/draft-gill-btsh-02),
checking whether the packet is actually in window, by ACLs that would
limit the impacts of spoofing from off path attackers (you can't target
my multihop infracture sessions from outside my network for example),
and by filters that would limit the sort of thing you could inject into
bgp (rendering prefix hijacking moot) ). I see broad evidence that MD5
values are extensively shared between sessions and effectively never
rekeyed (including cases where I've changed employers and the same asn
is using the same values for new peers). given the existance of
effective mitigations for the ibgp case, I've need seen a reason to
employ it internally or to explore support for rfc 4808 mechnisms since
key rolling is effectively an external coordination problem.
Due to window checking and the ttl hack, the best vantage point for
launching an attack against a single hop ebgp sessions is as an on path
attacker (such that you would be able to identify source port and
window), layer-2 exchanges which flood unicast traffic (a hub I guess or
any public exchange with broken mac learning) would seem particularly
vulnerable since there are many on path neighbors. That is no longer a
normal topology. 
we were unable to find bgp mib counters. there are igp interface
counters, but that was not our immediate interest. we did find
that md5 failures are logged.
I can't quite get there either.
md5 failures I see quite a lot of, as peers that formerly have it
configured fail either temporarily or over longer timescales. md5
failures for unestablished connections aren't very interesting in this
case.
I have thousands of establish connections that last a very long time at
public exchange points, so the threat of tcp rsts to sessions is clearly
not being realized.
my memory is that seq num guessing and sending rst was the core problem
motivating tcp/md5 for bgp, and btsh came some years later. but no big
deal.
i think that, indeed, md5 keys are shared across many links *within* an
op's infrastructure. but, since integrity, and not privacy, is the
goal, this does not seem risky. carrying keys to new networks seems a
bit risky as does re-use with multiple external parties.
given the existance of effective mitigations for the ibgp case, I've
need seen a reason to employ it internally or to explore support for
rfc 4808 mechnisms since key rolling is effectively an external
coordination problem.
if i need to roll keys on ibgp, i suspect i have a far more serious
problem than if it is ebgp, twice as serious at a minimum 
< rathole >
i am not much worried about a mesh which floods unicast. can you even
buy devices which support that any more? a while back, i had to really
dig in the closet to find one at 100mbps so i could shark mid-stream.
I have thousands of establish connections that last a very long time
at public exchange points, so the threat of tcp rsts to sessions is
clearly not being realized.
my theory is that, as the attacks were mitigated the attackers moved on
to other things. after all, the non-nuisance benefit i get by resetting
your bgp session with margaret is shifting your traffic past some place
i can mitm or to a more expensive, to you, link. the attackers moved on
to more lucrative endeavors.
randy
With regards to BGP, the MD5 thing was promulgated to counter what was a largely theoretical threat. iACLs, and later GTSM and CoPP and LPTS and so forth really obviated the need for it.
For IGPs, MD5 was belt-and-suspenders against someone deliberately or accidentally bringing up a new router and manipulating traffic internally. Passiving the IGP on non-core links was the BCP, but often was honored in the breach; pushing an additional feature for 'security' purposes got some folks' attention when the passiving BCP was ignored.
We still see DDoS attacks against routers, of course. But the goal there is disruption of availability, not trying to move traffic onto some alternate path which would somehow benefit the attacker.
< rathole >
i am not much worried about a mesh which floods unicast. can you even
buy devices which support that any more? a while back, i had to really
dig in the closet to find one at 100mbps so i could shark mid-stream.
I'm not actually worried about it because it is rare, and not a feature,
that said, unicast flooding is in fact something we detect on exchanges
with a fair amount of frequency e.g. 2-3 a week across the exchanges
were we are present. That traffic gets discarded on our ingress but you
can count dport 179 packets in there that aren't yours. I certainly
wouldn't build a business model around gaining insight from that
information leakage (and the bulk of the traffic is whatever the
neighbor is exchanging, with someone else, from looking at mac's that
sort of thing tends to be one sided unless for example it's a whole switch).
With regards to BGP, the MD5 thing was promulgated to counter what was
a largely theoretical threat.
the rst attacks were a very serious problem. attacks were very real and
very disruptive. gtsm et alia were a few years later.
We still see DDoS attacks against routers, of course.
i am focused on bgp, not the daily craptastic packet fling.
randy
Out of curiosity, are you asking for a specific research/project that you need some data for?
GTSM is not a replacement for the ACL filtering the bgp speakers or the MD5 ( that is widely supported).
If GTSM is not supported you can always predefine the TTL it in the session and manipulate the defaults ( 1 for EBGP 64 IBGP) yet this is works on small scale networks.
Another practice is to define who starts the session ( as port 179 is not hard to figure). Yet in all case the ACL is your first defense who is allowed for TCP and when the session is established what they can advertise and what you are willing to accept across the session ( NXT hop, n0. of routers, as-path, communities…etc).
https://www.noction.com/blog/bgp_security_md5_password_and_gtsm
|
BGP Security – the MD5 password and GTSM | Noction
www.noction.com
There are three security mechanisms that can protect against potential security issues with BGP: the BGP TCP MD5 password, IPsec and GTSM…
|
Brgds,
LG
* randy@psg.com (Randy Bush) [Wed 15 Aug 2018, 04:27 CEST]:
my memory is that seq num guessing and sending rst was the core problem motivating tcp/md5 for bgp, and btsh came some years later. but no big deal.
And a few looking glasses exposed detailed TCP window information when run against certain hardware vendors' routers, making that very easy.
-- Niels.
Nah, they aren’t asking about the other things, and only the order of operations which vary per vendor will matter.
If I am reading correctly, they aren’t asking about only successful MD5 attacks, but MD5 attacks in general.
All the rest of your listed security configurations would be ‘extra’ router demographics.
-Garrett