Large-scale IP network data analysis for anomalies detection thanks to SVM

Abstract

An SVM (Support Machine Vector) algorithm has been implemented to sense traffic anomalies through a largescale IP Network. We have applied this algorithm on data provided by the well-known large-scale American IP Network (Abilene Network). The developed SVM algorithm can classify the Network traffic into two cat egories of classes namely: normal; and abnormal. The implementation of this algorithm has been performed on real collected data thanks to Netflow protocol and has yielded satisfactory results with a classification rate going over 96% and a false alarms rate lower than 10%.