nanog mailing list archives

Re: DDOS Simulation


From: "Roland Dobbins" <rdobbins () arbor net>
Date: Thu, 30 Jul 2015 03:05:28 +0700

On 30 Jul 2015, at 2:38, alvin nanog wrote:

there is no need to pay people to attack your servers ...

Unless you don't have the expertise to do it yourself. Again, I advocate an organic defense capability and an organic testing capability, but there are many organizations which unfortunately don't have these, and they must start somewhere.

        - tcpdump and wireshark will tell you everything the attackers are
        doing to your network right now that needs to be defended against

On small, single-homed networks, sure. On networks of any size, this doesn't scale.

Flow telemetry scales.

if a mid-level wanna be attacker wants to target your servers, they're
just as equally easy to mitigate and prevent and probably sending you
100,000 "ddos packets" per second because they can ( bigger zombie network :-)

100kpps is nothing. Of course, so many servers/services are so brittle, fragile, and non-scalable that most DDoS attacks are overkill by orders of magnitude.

if you are being targeted by "masters of deception" you have no solution other than get local law enforcement involved to track down the originating
attackers

I'm not sure who or what 'masters of deception' are in this context, but attribution has nothing to do with DDoS defense.

Defending against serious attackers with lots of resources is taking place every minute of every hour of every day. There are many techniques and tools available, most of which have been discussed multiple times on this list over the years. Here's one such example:

<http://mailman.nanog.org/pipermail/nanog/2010-January/016747.html>

all ddos mitigations is almost 100% guaranteed to fail a volumetric
DDoS attacks ....

This is incorrect.

the DDoS attackrs probably have access to a bigger zombie
network than most major corp ...

This is true, in many cases - and is also not an issue for properly-provisioned, coordinated DDoS defense mechanisms and methodologies.

the attackers job is not to get caught and
is not ez to be hiding if law enforcement wanted to catch them :-)

Again, attribution is a completely separate issue.

        nping "send 100,000 packets/sec" x 65,000byte/packet  192.168.0.0/16

FYI, 'line-rate' for 64-byte packets at 10gb/sec is ~14.8mpps.

by the same premise, if i had to pick ONE ddos mitigation strategy, i'd
tarpit all incoming TCP-based ddos attacks which should crash the
attacking zombie server under sustained tcp-based ddos attacks

There is no one tactic (this is not a strategy) which can be picked, as any kind of traffic can be used for DDoS attacks. With regards to TCP-based attacks, it's a subset of those which are connection-oriented and are thus susceptible to tarpitting-type techniques.

-----------------------------------
Roland Dobbins <rdobbins () arbor net>


Current thread: