How an attack works
The attacker controls a network of infected devices – computers, cameras, routers. On command they all send requests to your address at once. Each request on its own looks normal, but together they exhaust the server’s bandwidth, memory or processor.
Three main types
- Attacks on the link: more data than the connection can carry. The server may be perfectly healthy, but nobody can reach it.
- Protocol attacks: requests that force a server or network equipment to hold thousands of unfinished connections.
- Application attacks: plausible requests to search, sign-in or basket pages that are expensive to process.
Different types need different protection, so there is no single solution “against everything”.
How to tell you are under attack
The service suddenly slows down or stops responding, although nothing was updated; the logs show a sharp rise in requests from many addresses; the provider reports that the link is overloaded. You need monitoring that alerts the person on duty before customers start calling.
What to do in advance
- List the services whose outage would cost the most, and decide how much downtime each can tolerate.
- Agree with your provider on what they do during an attack on the link, and have a backup link from another provider.
- Protect the entry point: requests without the right access should be dropped before any expensive processing.
- Have backup entry points in different networks that your own users can switch to.
- Prepare team communication that does not depend on the services under attack.
Why separate communication matters during an attack
An attack is often used as cover: while everyone is busy with the unavailable service, the attacker tries another way in. If email and the work chat run through the same node that is under attack, the team is left without communication exactly when it is needed most.
Third-party scrubbing services: pros and cons
A common solution is to route traffic through an external service that filters out the attack. This works well for large attacks on the link, but all your information passes through someone else’s infrastructure. For public authorities and closed networks this is often unacceptable.
What to do when an attack is already under way
- Inform the person on duty and the manager over a channel agreed in advance that does not depend on the service under attack.
- Contact your provider and check whether the link is saturated.
- Move your users to the backup entry points.
- Watch for other events on the network: the attack may be cover.
- After the attack, write down what worked and what did not, and update the plan.
How CUBRA approaches this
Our own DDoS protection is built into the CUBRA Node: unwanted requests are filtered out before processing, and under attack your own apps switch to new entry points by themselves. Traffic is not handed over to a third-party service. If an attack saturates the provider’s link completely, no node inside the network can change that – so we plan backup links together with you. Read more on the DDoS protection page (switching entry points under attack — in development).