RSU based Joint Congestion-Intrusion Detection System in Vanets Using Deep Learning Technique

Abstract

Vehicular Ad hoc Network (VANET) is a technology that makes it possible to provide many practical services in intelligent transportation systems, but it is also susceptible to several intrusion threats. Through the identification of unusual network behavior, intrusion detection systems (ID Ss) can reduce security vulnerabilities. However, rather than detecting anomalous network behaviors throughout the whole VANET, current IDS systems are only able to do so for local sub-networks. Hence there is a need for a Joint Congestion and Intrusion Detection System (JCIDS). We designed an JCICS model that can collect network data cooperatively from vehicles and Roadside Units (RSUs).This paper, proposes a new deep learning model to improve the performance of JCIDS by using k-means and a posterior detection based on coresets to improve the detection accuracy and eliminate the redundant messages. The efficacy of the current Recurrent Neural Network (RNN) and Honey badger Algorithm (HBA)on the fundamental AODV protocol is combined with the advantages of the JCIDS is suggested in this protocol. First, formation of clusters using vehicle’s mobility parameters like, velocity and distance to enhance route stability. Moreover, a vehicle will be chosen as Cluster Head with highest route stability. Second, the efficient intrusion detection is achieved with the consumption using RNN method. In the RNN, the optimal weighting factor is selected with the help of HBA. The RNN is performing efficient prediction with the assistance of HBA. The finest path for data dissemination is selected by choosing link lifetime, hop count and residual energy along the path.As a result, multimedia data streaming is improved network life time, in terms of reduced packet loss ratio and energy consumption as compared to existing DNN and SVM scheme for different node density and speed.