Classification of network traffic in LAN

Abstract

Classification of Network Traffic is one of most important issue in network management and detection of Intrusion attacks play a vital role in it. To have a holistic picture of the network intrusion detection, selection of appropriate feature is very important; it reduces analysis effort and time too. Data mining can be very fruitful for feature selection and intrusion detection. In this paper, Tcpdump is used to capture network traffic and visualize different set of features using k-mean clustering. KDD'99 corrected intrusion detection dataset is evaluated to find out most important and relevant features and an algorithm based on the features is proposed to detect different types of dos, probing, u2r and r2l attacks with an accuracy of more than 80%.