International Journal of Computer Networks and Applications

Number V-3-I-6, 2016

129-138
Enhanced Route Discovery Mechanism of Ad-Hoc On Demand Distance Vector for MANET

Authors: May Sayed A. Nouh, Salwa H. El-ramly, M. Zaki, Husein A.A Elsayed

Abstract

View in PDF

Number of views: 327

The Ad-hoc On Demand Distance Vector (AODV) is one of the most famed reactive routing protocols, which is used in Mobile Ad Hoc Networks (MANETs) for routing purposes, where the routes are built only when it is required. The AODV is related to the problem of route constructing and maintaining among any two-node in the presence of the dynamic topology of a network. High routing control packets, during route construction, is a limiting factor of AODV. Also, the route created during route discovery process may not be optimum (in terms of the number of hop counts between the originating and target node). The construction of non-optimal route arises from many reasons such as temporary congestion of one or more links in the short path to the target node, or nodes in those paths may contend the channel late, or the nodes may be busy for that moment to entertain other engagement. Moreover, during the route discovery phase, a large amount of unnecessary control packets may be generated which result in the debasement of the performance and the efficiency of the protocol by congesting the network and increasing the overhead. In this article the cases at which non-ideal routes are created and unimportant control packets generated during route discovery process will be identified. The modified AODV (AODV_MOD) with enhanced route discovery mechanism is proposed that will be used to avoid these cases by suppressing unimportant control packets and avoiding non-ideal routes formation. Simulation results, conducted in network simulator (NS2), prove the skillfulness of the proposed enhanced route discovery scheme from point of view of packet delivery ratio, end-to-end delay, and normalized overhead.

139-147
Usage of Machine Learning for Intrusion Detection in a Network

Authors: Prachi

Abstract

View in PDF

Number of views: 299

Increase in volume and intensity of network attacks, forcing the business systems to revamp their network security solutions in order to avoid huge financial losses. Intrusion Detection Systems are one of the most essential security solutions in order to ensure the security of any network. Considering huge volumes of network data and complex nature of intrusions, the performance optimization of Network Intrusion Detection System became an open problem that is gaining more and more attention from the researchers nowadays. The objective of this paper is to identify a machine learning algorithm that provides high accuracy and real-time system application. This paper evaluates the performance of 15 different machine learning algorithms using NSL-KDD dataset on the basis of false discovery rate, average accuracy, root mean squared error and model building time. Firstly, 5 machine learning algorithms out of 15 are chosen on the basis of maximum accuracy and minimum error in WEKA. Simulation of these machine learning algorithms is performed using 10-fold cross validation. Thereafter, the best machine learning algorithm is selected on the basis of maximum accuracy and minimum model building time so that it can be readily implemented in real-time Intrusion Detection Systems.