Modeling of Support Vector Machine for Intrusion Detection System in Ad-hoc Networks Using R Programming

With the emerging demand of ad-hoc networks which are self-deployable and infra-structureless in nature, security of ad-hoc networks has become a vital issue. In ad-hoc networks, a challenging security problem is the presence of intruders. These intruders can disrupt the network traffic, choke bandwidth, cause network congestion, and consume limited resources of nodes by making byzantine type attacks. Modeling of a system that detects such intruders in the network has become a hot research area. In this paper, we propose a framework for such an innovative system to detect intruders in the network with relatively low false alarm rates. Novelty of our paper includes that we propose a framework that has two modules. In the first module, we design a simulator in C++ to simulate behavior of nodes and to monitor as well as capture details related to parameters of nodes and traces of packets sent or received by nodes in networks. This simulator uses a statistical decentralized approach for creating a labeled data set. In the second module, we model a comprehensive security scheme using R programming that models pre-processing, detection, and response system by which such intruders, exhibiting byzantine type attacks, can be detected with high accuracy. Our framework is based upon a supervised classification technique in machine learning. It implements a support vector machine (SVM)-based intrusion detection system and uses a labeled data set, recorded by our simulator, as training data set for the learning of the model/framework. In this paper, the performance of our proposed framework that is based on anomaly detection technique is evaluated and the results indicate that the proposed framework exhibits high accuracy and low false alarm rate for detecting the intruders in ad-hoc networks.