Restricted Boltzmann Machine based detection system for DDoS attack in Software Defined Networks

Abstract

Software Defined Network is an innovative network architecture which provides network control through software logic. It decouples control and data plane to customize the network according to the user needs. OpenFlow, a standardized network protocol acts as an interface between controllers and switches. The softwarized controllers are highly vulnerable for Distributed Denial of Service attacks. The proposed detection system uses an unsupervised stochastic Restricted Boltzmann Machine algorithm to self-learn the reliable network metrics. This algorithm detects and classifies the type of DDoS attacks in a dynamic network environment by framing a new context. The results prove that RBM based DDoS detection system achieves higher accuracy than the existing methods.