DDoS Update

  • 🐕 I am attempting to get the site runnning as fast as possible. If you are experiencing slow page load times, please report it.

Null

Ooperator
kiwifarms.net
Joined
Nov 14, 2012
I've taken some measures against DDoS attacks. I'd like to explain what a DDoS is and how it affects services, and what has been done to help protect against them.

The most simple form of attack is called a Layer 7, or application-level "slow attack". It relies on exploiting the victim's resources. An attacker serves a large number of requests that the server works very hard to respond to. The result is overworking itself to the point of not being able to serve real visitors.

Attack 1.png


To protect against these sorts of attacks, we employ Cloudflare. Cloudflare is a massive application that sits between the request and our servers. They absorb malicious traffic and help filter out attacks. Depending on the cleverness of the attack, sometimes Cloudflare has to respond by being more restrictive. This is when you see things like browser checks and CAPTCHAS.


Attack 2.png


Cloudflare works in a particular way. Clever people can find ways around it. Some applications are vulnerable to identification attacks, that reveal the true source of the servers. With this information, attacks could potentially bypass our front line of defense.

To handle this, I have done two things. The server now throws away any connections not from Cloudflare. That way, a direct Level 7 attack will not work. Additionally, vulnerabilities have been moved to a smaller and less important Linode. If that server is attacked, no one will care.

Attack 3.png


Since attackers will be forced to use Cloudflare to even talk to the forum, they will have their options limited. Attacks can still work, depending on what methods they use. To help combat this, I've set up three different domain names. If one of the domains is attacked, I close it and will let people know what other domain to use for the time being.

Attack 4.png


That's a very inconvenient thing to do, so it's a last resort. Still, the option is there.

Most bases are covered. Uncovering the forum's true source won't help much, and is going to be very hard to do. Attacking Orange (the sub-server) won't bring down the website. Attacking Cloudflare is almost completely useless. Even if we can't stop attackers from breaking through Cloudflare, I can shut them off like valves to keep the server online (at the cost of annoying people).

Is there a weakness? Of course. Everyone reading this already knew what a DDoS was, or at least had heard of them. That's because the Internet is very vulnerable to this sort of attack. It's a big deal. That's why it's a felony in many western countries, including the United States.

A large-scale botnet attack with thousands of infected computers directly hitting the server would look something like this.

Attack 5.png


At this point, there's nothing to be done. An attack of this magnitude shakes the very foundation of the world wide web to its bones and reveals the innate flaws of its being. The best I could do is explain to my service host the situation. I could obtain a new IP, or they could blackhole any non-Cloudflare IPs (like with what I've done on a server level).

The blue bubble in my earlier diagram ceases to work at this point because, by the time the traffic reaches the server, it's already done its damage. Compared to a Level 7 attack each connection is massively inefficient, but because there's so many of them, it doesn't matter. The Internet, as we all know, is a series of tubes. This sort of massive attack will clog them no matter what sort of fancy garbage disposal you have.

A botnet attack like that costs about $200 dollars a day, from what I've seen.
 
Last edited:
have we identified who is trying to attack us, to any extent?

i believe your methods will indeed likely assist in protecting the forums, but since i believe a large-scale botnet is unlikely (because only the most advanced malware-creator would likely have access to such a group, and anyone angry at these forums likely has no such access) then we likely don't have to worry about more violent attacks. plus, a botnet infected by malware would have already been destroyed, as far as i know. we would be grossly aware of such a piece of code (or probably those would lurk on antimalware blogs would.) so a botnet attack is unlikely. i have heard of people handing money to those who create botnets, and control them. that's the only real way a botnet like that would basically attack, as far as i know. this is extremely unlikely.

knock on wood in case though.
 
Last edited:
I've answered this question at least 5 times in public.

Yes, I know who it is.
Yes, I know why.
No, I'm not making it public. If he/she/they/xir want to take credit, they can.
that's fine. since DDoSing is a felony i, and probably a lot of kiwis here, can understand why you wouldn't want that info out.
 
I disabled DDoS protection for a second to fix a caching error with Cloudflare and immediately the server dropped. Normally, the attacks only happening during events. This is the first time it's gone on for more than an hour.

Buckle down, my Kiwis, this is the long haul.

As an immediate fix, I've forced stylesheets based on domain.
 
You guys misunderstood me. $200/day is the "DDoS Deluxe" package in the last figure. The one that would completely obliterate everything.

More likely prices for attacks that we are currently able to endure are seen in this article:
http://www.computerworld.com/article/2837175/symantec-sees-rise-in-high-traffic-ddos-attacks.html

f41voef.jpg


I'm not super knowledgeable about the actual logistics behind orchestrating an attack. I just know how they work, and thanks to some reading I know a little bit about stopping them.

It's possible the attacker owns their own small botnet and are carrying out the attacks themselves. It's possible they are paying. My guess is that they are. I'm actually very suspicious that the person bought a service similar to the very one I've linked, because:


Nov 23th, we experience three separate attacks of escalating intensity. One during the Jace stream, one immediately after, and one much later in the afternoon.

Nov 24th, we experience an attack late at night.

Nov 25th, we experience a much longer attack during the Jace stream.


I'm thinking that, what they're doing, is buying the very short timeframe attacks.

Day 1, $4 for Package A 1 hour, $6 for Package B 1 hour, $15 for Package C 1 hour. Each attack got much harder to deal with. $25 total.

Day 2, Package C 1 hour. $40 total.

Day 3, Package C 4 hours. $70 total. This is also the first attack that lasted much longer. Hours later I went to fix a caching issue by messing with Cloudflare and immediately the server went down again because the attack was still going on.


Now again, this is all crapshooting. I have no idea. If it's just a few friends attacking, maybe they could achieve these results? It just doesn't seem likely. Even if they did buy an attack, I don't know if this is the pricing. Again, I've seen everything from $5/day to $1000/day, based on the attack.

I'm certain they don't spend too much money, because we're still here, but I'm also doubtful they're carrying out the attack with personal devices. If they were, it wouldn't be over ever. It would be a perpetual siege. Money is more fleeting than time, however, so it would make sense to pick your attacks wisely if you had to pay for it.


Again, can't be certain of the costs, but this is what I know.
 
$200 a day to keep a bunch of autists from talking about other, slightly worse autists. That's Charb-level financial strategy, right there.

Well shit, people are actually paying $200 to attack our forums? All we do is talk about online weirdos, no idea why they feel so strongly that they need to stop us.

EDIT: Okay, so they're probably not spending $200. Still, spending any amount of money to attack our forums is just silly.
 
Back