Linked by Thom Holwerda on Thu 28th Mar 2013 00:36 UTC, submitted by MOS6510
Internet & Networking "The New York Times this morning published a story about the Spamhaus DDoS attack and how CloudFlare helped mitigate it and keep the site online. The Times calls the attack the largest known DDoS attack ever on the Internet. We wrote about the attack last week. At the time, it was a large attack, sending 85Gbps of traffic. Since then, the attack got much worse. Here are some of the technical details of what we've seen."
Thread beginning with comment 556952
To view parent comment, click here.
To read all comments associated with this story, please click here.
Laurence
Member since:
2007-03-26

t doesn't seem like source interface filtering is a great solution to me because on the internet there's technically no requirement that packets come in from the same interface they'll return out of. In multi-homed setups this can even be explicit. Load balancers might do the same thing. But even in other less exotic cases internet routers can switch paths dynamically as they rerun the shortest path algorithms, I don't know just how frequently this happens, but it's the reason UDP packets can arrive out of order.

For core switches, you'd be right. But from what I've read, that method could work for routers on the edge of networks. But that's just what I've read, you might well be right ;)


So do you agree that source interface filtering could negatively affect legitimate users?

My guess is it would either work well or not at all. I'm by no means a networking expert though so I'll have to take the lead from someone else.


It's a DNS problem, so I feel that a DNS fix should be used instead of modifying our routers. It's much easier to update dns software than a router. My understanding is that many commercial routers achieve their performance in hardware and become underpowered if too many packets get tossed around into the software stack.

I'd argue it's more a problem with the UDP datagram than DNS specifically. DNS just exposes that weakness of UDP. So if we just fix DNS then I'm sure someone will find another UDP service that can be exploited in the same way (possibly games servers?)

Edited 2013-03-28 16:35 UTC

Reply Parent Score: 2

Alfman Member since:
2011-01-28

Laurence,

"I'd argue it's more a problem with the UDP datagram than DNS specifically. DNS just exposes that weakness of UDP. So if we just fix DNS then I'm sure someone will find another UDP service that can be exploited in the same way (possibly games servers?)"

I guess you could put it that way. It's true that UDP does nothing to confirm the sender IP, but to fix it at this level would mean converting UDP to a stateful protocol with a bidirectional handshake.

Between all the IP protocols already invented, we probably already have something that would work, but it does little good until these actually saw widespread support.


http://en.wikipedia.org/wiki/Reliable_User_Datagram_Protocol

http://en.wikipedia.org/wiki/Stream_Control_Transmission_Protocol

So, the easiest fix in this case is probably just running DNS over TCP.

Edited 2013-03-28 17:56 UTC

Reply Parent Score: 2

Laurence Member since:
2007-03-26

I did read someone else suggesting that and at the time I didn't take their suggestion all that serious because of the disruption it could cause. But thinking about it again, it's probably a good long term goal.

And with that, I think you're probably right that the best solution is at the name server end rather than trying to patch all the edge routers.

It's been an interesting discussion this. Thanks for your insights ;)

Reply Parent Score: 2