Design of Network Architecture for Intrusion Detection Using Spanning Tree Multiclass Classifier in Manet

Security in mobile ad-hoc network plays a strategic role to ensure high level of protection without any intrusions in computer networks. Most of the intrusions in mobile ad-hoc network are traced and detected by collecting traffic information and classified according to different classification algorithms. With individual traffic classifiers design, packet delay is expected to surely go up with overhead cost rate. These classification processes hosted on mobile ad-hoc network failed to develop multiclass classifier system under varied conditions and minimized the efficiency on detecting intrusions in network architecture. To present mobile ad-hoc network architecture, with multi-classifier intrusion detection system, a mechanism called, Tuning Spanning Tree Multiclass Classifier (TSTMC) is proposed in this paper. The TSTMC mechanism works with multiple classification process in MANET to detect the intruded distribution nodes. First, Multiclass Neural Shannon’s Entropy based approach analyzes the network traffic properties based on the count of source destination pair, packet type, packet size and payload. After analyzing the traffic rate on MANET, the result are used in the next step to construct spanning tree network architecture. Secondly, the spanning tree is constructed with Tuning Tutte polynomial operations where the internal and external traffic over mobile ad-hoc network is classified. Tuning Tutte polynomial operation is used effectively to detect the abnormal activities through tuning spanning tree. Analysis is accomplished for different parameters, including the maximum node velocity and the average multiclass error rate they experience, node density and their true positive rate on detecting abnormal activities, packet delay and classification sensitivity rate. Simulations using NS2 were used to measure the performance of the mechanism to compare it with the performance of two other intrusion models for mobile ad-hoc networks, namely, GP and SSUM. The measured results signify the effectiveness of the proposed mechanism in terms of achieved multiclass error rate, classification sensitivity and low packet delay.