Intrusion Detection in Manet Through Machine Learning Approach

Abstract

Mobile ad hoc networks (MANETs) have evolved into a leading multi-hop infrastructure less wireless communication technology where every node performs the function of a router. Ad- hoc networks have been spontaneously and specifically designed for the nodes to communicate with each other in locations where it is either complex or impractical to set up an infrastructure. The overwhelming truth is that with IoT emergence, the number of devices being connected every single second keeps increasing tremendously on account of factors like scalability, cost factor and scalability which are beneficial to several sectors like education, disaster management, healthcare, espionage etc., where the identification and allocation of resources as well as services is a major constraint. Nevertheless, this infrastructure with dynamic mobile nodes makes it more susceptible to diverse attack scenarios especially in critical circumstances like combat zone communications where security is inevitable and vulnerabilities in the MANET could be an ideal choice to breach the security. Therefore, it is crucial to select a robust and reliable system that could filter malicious activities and safeguard the network. Network topology and mobility constraints poses difficulty in identifying malicious nodes that can infuse false routes or packets could be lost due to certain attacks like black hole or worm hole. Hence our objective is to propose a security solution to above mentioned issue through ML based anomaly detection and which detects and isolates the attacks in MANETs. Most of the existing technologies detect the anomalies by utilizing static behavior; this may not prove effective as MANET portrays dynamic behavior. Machine learning in MANETs helps in constructing an analytical model for predicting security threats that could pose enormous challenges in future. Machine learning techniques through its statistical and logical methods offers MANETs the learning potential and encourages towards adaptation to different environments. The major objective of our study is to identify the intricate patterns and construct a secure mobile ad-hoc network by focusing on security aspects by identifying malicious nodes and mitigate attacks. Simulation-oriented results establish that the proposed technique has better PDR and EED in comparison to the other existing techniques.