DDoS Detection System Using Wavelet Features and Semi-supervised Learning

Abstract

Protection of critical information infrastructure is a major task for the network security experts in any part of the globe. There are certain threats that will never evade away despite sophisticated advancements in defense strategy. Among them, Distributed Denial of Service (DDoS) attacks have witnessed continual growth in scale, frequency and intensity. The impact of DDoS attacks can be devastating such that it creates severe ripples to the cyberworld. Nowadays, attackers are advancing towards different variants of DDoS attacks to escape from the detection mechanisms. In this paper, a new DDoS Detection system is proposed. Initially, wavelet based features are extracted and classified using semi-supervised learning to detect the DDoS attacks. Different wavelet families are studied and the combination of them seems to be robust and efficient and hence used as features. Machine learning algorithms are highly appreciated in many classification problems. There is a considerable demand for labeled dataset and hence to bridge the gap between them and unlabeled dataset, semi-supervised learning algorithm is employed to classify the attack from normal traffic. Extensive analysis is performed by conducting experiments and by using real-time dataset. Results obtained are convincing and hence can be modeled for real-time approach.