The goal of attacks on CPU time is to reach a 100% load of CPU server. One of the options is to process a large amount of information from a database, which takes a lot of effort.
A great example of an attack on CPU time is a SSL Renegatiation DoS attack. To perform it a client repeatedly connects with the server applying SSL protocol, whereas the server needs 15 times more processing power than a client to process the request.
How to protect the server from this attack?
First of all turn off SSL Renegatiation mechanism. Attacks on CPU time are very dangerous for all kind of services and result in serious damage. A software installed on you server and operating as a filter against potentially dangerous request can be a good security measure.
The goal of attacks on server`s memory is to fill the internal memory or a hard disk, or a critical part of memory completely. When the goal is reached the server is unavailable. The classic example of the attack is a TCP SYN flood attack. It`s main objective is to create multiple half open-connections to limit the access to the server. To reach the goal the attacker sends a number of TCP – segments with mounted SYN flash.
If your set up is correct, the information about new connection will be stored in the server`s response, that sends to the initiator of connection. When the server gets the client`s response it checks the information about the connection in the package. This security technology is called syncookies. It extracts the IP-address and the port from the income package, transforms the data into numbers and adds them to the Sequence Number + 1 field.
TCP FIN flood attack is similar to SYN flood attack but creates a number of half-open connections. In case like this, when the connection is established, a TCP-segment with FIN flag is sent to complete it. The server responds with FIN + ACK flagged segment and waits for the approval. If the approval is not send and the memory is filled. To fix the problem you need a patch from a product owner.
“Slow” attacks on HTTP protocol are performed in a bit different way. Apache web-server is most vulnerable to them.
The attacker acts in the following way: connects to a server and starts accepting and sending the data. He performs multiple connections to reach Denial-of-Service. As a result, server`s memory is filled due to the great number of data streams from the Apache server.
To protect the server you need to define the minimal and maximum time limit for sending the request and accepting the answer. If the client exceeds the limit the connection stops. Use specially created modules to set up the protection.
In the end, I would like to say that automatic systems are really important today. They manage various critical processes. Denial-of-Service attacks can lead to unexpected results, thus the issues of protection against DoS/DDoS attacks are as critical as never before and won`t be minor in future.