Classification and Prediction of traffic in Optical Burst Switched Networks using Machine Learning

Abstract

   Optical burst switched (OBS) networks has been emerged as a new infrastructure of this century for optical internet. The classification of network traffic and segregation of normal traffic from the malicious traffic is vital for security and data integrity in OBS. The traffic classification mechanism should be dynamic and capable enough to segregate the network traffic in a quick manner, so that the malicious traffic is identified, then deflected at the early stage and the normal traffic is to be channelized to the destined nodes. In this paper, we are presenting two phased machine learning based mechanism. The first phase focuses on the segregation of network traffic into four different classes which assist in reducing the congestion over channels, enhancing the network throughput and providing secure services to end users. The second phase is providing insights into the prediction of traffic load using convolutional neural networks for channelization of the normal traffic over under-utilized and least loaded channels.