A denial-of-service attack takes no data. It makes the service unreachable, which is why many organizations give it lower priority. For a webshop or a customer portal, though, downtime means lost revenue straight away, and more and more often it comes with an extortion demand.
Three types
Volumetric. Saturates the bandwidth with a large volume of traffic. This is the most common type, and you cannot defend against it on your own infrastructure: by the time the traffic reaches you, it is too late. Only your provider or a specialized service can filter it.
Protocol-based. Exhausts state tables instead of bandwidth, with half-open connections and fragmented packets. The firewall or the load balancer runs out of capacity.
Application-layer. Uses a small amount of traffic to load up a resource-intensive operation such as search, report generation, or login. It is hard to tell apart from legitimate traffic because it consists of technically valid requests.
What your provider can do
This is the first thing to clarify, and most organizations do not know the answer.
- Does your provider contract include DDoS filtering at all?
- Does it switch on automatically, or do you have to request it?
- Up to what traffic volume does it protect?
- What is the response time, and can you reach them around the clock?
If you are unsure of any of these answers, don’t find out during an attack.
What you have to handle yourself
Application-level rate limiting. Per-user and per-IP rate limits on resource-intensive operations (search, export, login). This is the only protection against an application-layer attack.
Caching. Anything static should not be served by the application. A well-configured CDN also absorbs a large part of a volumetric attack.
Resource limits. Connection counts, request sizes, timeouts. If these are not set, even a single slow client can tie up resources.
Autoscaling, if you have it. In the cloud, scaling absorbs the peak, but without an upper limit the attacker is running up your bill.
What to prepare in advance
Contact list. Who to call at the provider, on which channel, and the contract number. Keep a printed copy as well.
Decision points. When filtering gets switched on, when features get restricted, and when you start communicating with customers.
Static fallback page. If the application cannot handle the load, a simple cached page that tells people what is going on is worth more than an error message.
Communication template. What you tell customers, and where. If your own site is unreachable, you need another channel too.
What not to do
Don’t block IP ranges blindly. During an attack it is easy to lock out real customers by accident, and that is hard to spot afterwards.
Don’t assume the firewall will take care of it. In a volumetric attack, the firewall itself becomes the bottleneck.
Don’t skip the rehearsal. If nobody has ever walked through the DDoS plan, you will find out what is missing from it during the attack. We covered this in our article on the first 24 hours after an incident.
Our IT security services also include a review of your availability architecture.