Day 6: Most common attacks on SDN controller networks

Today's post let's look at what security issues we can resolve in the architectures we looked at yesterday. The most common malicious attacks that can happen in a distributed SDN controller environment are:

  • DoS attacks (Denial of Service attacks): In these attacks many request messages are sent from the same host to the server (here, controller). The number of packets increases to an extent that the traffic it causes utilizes all of the resources on the server side and thus leads to the system crashing. Thus, to protect the controller network from these attacks, we can observe the entropy of few of the fields of the packet header. To name a few, it could be source IP address, source MAC, destination IP and destination MAC. We can establish the entropy levels which convey that the attack has started.
  • DDoS attacks (Distributed Denial of Service attacks): This is very similar to the above, the only difference being that instead of one host bombarding the request packets, a set of hosts do the same in a more distributed fashion. The complexity of detection increases further since the source IPs and MACs will be different here. This may lead to false positive errors while detection. But there are few thresholding mechanisms to prevent the same. We shall look into this later.
  • Port Scan attacks: This is also a malicious attack where a host can bombard the server side with many request packets but by varying the destination port. These attacks have the potential of revealing the different processes and services the server is running. The attacker can also get an idea of the operating system that the server is running. Such information in the bad hands can land us into much trouble and hence addressing these attacks as well becomes critical.
I have given the gist of all possible attacks we would be addressing in the further posts. Deciding on the architecture is the next big step. We are working on the same .

Image result for work in progress sign
 
Watch out for the architecture we come up with, in tomorrow or day after tomorrow's post.

Comments

  1. In today's world, an attacker will not use its own IP Address as source. It will camouflage its src IP Address. This will work for IP, and UDP but not for TCP though. Thus, most attacks for DDoS.

    ReplyDelete
    Replies
    1. Agreed sir. After considering your inputs, we thought of concentrating on Syn flood and Smurf attacks. Once we have successfully built solutions for them, we can hopefully look into other variants. But for now, we have not kept that in the scope of the project.

      Delete

Post a Comment

Popular posts from this blog

Day 12: Master Slave SDN Controller Architecture

Day 1: Understanding Ransomware and how to detect them?

Day 50: Tcpreplay and tcpliveplay approach