Bi directional sparse attention recurrent autoencoder based intrusion detection for VANET security with tuna swarm optimization

Vehicular Ad-hoc Networks (VANETs) have garnered a lot of consideration and research in last decades. By enhancing safety and comfort, VANETs are essential to the development of self-driving and semi-self-driving cars. Security risks that are specific to VANET or that are observed in ad hoc networks, however, pose significant difficulties. External communication with the situation through data control and Cooperative Awareness Messages (CAMs) exchanges is crucial for these vehicles. Numerous assaults could possibly target VANETs. VANETs may be vulnerable to several types of assaults. The lack of intrusion detection systems, inadequate encryption methods, insecure communication routes, and inadequate authentication processes are some of the technological flaws in VANETs that make smart vehicles vulnerable to attacks. The security and effectiveness of smart vehicle operations suffer because these flaws allow attackers to intercept, change, or interrupt communication between vehicles. Owing to these security concerns, this study introduces a Bi-directional sparse Attention-recurrent Auto Encoder (BSAR-AE) based Intrusion Detection System (IDS) in VANET. The system uses the Tuna Swarm optimizer (TSO) with the enhancement through the implementation of Deep Neural Network (DNN) based feature selection in VANET. It is a useful technique for resolving issues with deep layer neural networks, like low velocity and overfitting issues in learning. The evolutionary based optimization effectively tuned the hyper parameters of the DL model which will improve the detection accuracy. The system use of the CICIDS2017 dataset, which is associated with actual data and the experimented analysis shows that the proposed model is effectual for detecting the malicious attacks in VANET with improved accuracy of 98.7%.