Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Information Technology Department

Includes bibliography (sheets 60-65)

ABSTRACT
Over the years, the advent rise of cloud computing and how it has become advantageous to organizations that are currently using cloud computing services to enable them to service their clients with easy access to services, scalability of infrastructure, cost reduction, performance increase, flexibility and enabling a better streamlining of processes. With all these great advantages of the cloud computing model, there are bound to be disadvantages also which brings us to one of the biggest challenges of the cloud computing service model which is security, but this service model has some great features like CPU Utilization, ability to auto scale and elasticity of the infrastructure which already helps adequately in cost reduction and furthermore, increases the use of resources. These features can also be a disadvantage if illegitimate clients use these features against the cloud computing service model resources by propagating a denial of service attack which can then be translated to a different but similar form of attack on the cloud resources by continually sending requests to the servers and abusing the ability of the service model to be elastic and autoscale in turn creating cloud service cost to increase exponentially. This form of attack is called Economic Denial of Service.
This study is going to focus on some already existing economic denial of service mitigation techniques and the most common faults they have. This study is also going to also propose a new mitigation technique that will be implemented at the cloud provider level considering permission to access cloud resources per client based on giving parameters like; the trust factor of the client, the maximum requests per second received from a client and the ability to prove a client is not a bot/zombie by implementing a graphical Turing test. The proposed technique will first look at the ability of the model to auto scale giving a set threshold to confirm if an attack is in effect, if it detects an attack, the cloud computing system then tries to check if it's coming from bots or a legitimate client. Once the attack source is confirmed, all illegitimate traffic/requests will be dropped. Using CloudSim as the simulation environment shows that the proposed mitigation technique can differentiate between legitimate and illegitimate users. This study will also compare the proposed approach to the well-known EDoS-Shield mitigation technique.
Keywords: EDOS, Graphic Turing Test, Load Balancer, NAT, IP, Scaling-Up, Scaling-Down, Threshold, Virtual Firewall, Virtual Node, Virtual Machine