Adaptive resilience for computer networks: Using online fuzzy learning
Azman Ali, David Hutchison, Plamen Parvanov Angelov, Paul Smith · 2012
Challenges on the Internet infrastructure such as Distributed Denial of Service (DDoS) attacks are becoming more and more elaborate. DDoS attacks have emerged as a growing threat not only to businesses and organizations but also to national security and public safety. DDoS attacks have become more dynamic and intelligent than before, prompting equally advanced responses for dealing with these attacks. In this paper, we aim to apply a modified version of our “D2R2+DR” Resilience Strategy to help solve this problem. We adopt a fuzzy rule based learning technique with self-evolving capability. Two scenarios are proposed for our case studies. The first is focusing on the traffic classification problem, while the latter aims to capitalise on policy to control the activities within the resilience framework. The two scenarios are investigated using a resilience simulator framework that provides a near-realistic large scale attack scenario on a network operator network topology. The novelty of our proposal lies in the combination of policy-based management and the application of an advanced online, self-evolving learning technique that has been proved to work in uncertain environments or in environments where obtaining detailed knowledge from the surrounding is almost impossible. It also requires fewer computational and human resources, which is key to contributing towards a more practical, resource efficient and autonomous resilience framework.