Artificial Intelligence based Intrusion Detection System to detect Flooding Attack in VANETs

Abstract

Vehicular Ad hoc Networks (VANETs) are classes of ad hoc network that provides communication among various vehicles and roadside units. VANETs are decentralized and due to this, these are susceptible to many security attacks. These attacks mainly affects the five requirements- the availability, confidentiality, integrity, non-repudiation and authenticity of the system. Intrusion Detection System (IDS) are used to combat these attacks. Flooding attack is one of the major security threats to VANET environment. The current thesis proposes a hybrid Intrusion Detection System which improves accuracy and other performance metrics using Artificial Neural Networks as classification engine and Genetic algorithm as optimization engine for feature subset selection. These performance metrics has been calculated in two scenarios namely misuse and anomaly. Various performance metrics are calculated and compared with other researchers work. The results obtained indicate high accuracy and precision and negligible false alarm rate. These performance metric are used to evaluate intrusion system and compared with other existing algorithms. The classifier works well for multiple malicious nodes. Apart from machine learning techniques, effect of the network parameters like throughput and packet delivery ratio are observed.