Brain Storm Optimization with Deep Learning-Based Intrusion Detection System in Vehicular Adhoc Networks

Abstract

Vehicular adhoc network (VANET) is an empowering technology in recent transportation systems for offering valuable information and safety, but prone to several attacks, such as active interference and passive eavesdropping. Intrusion detection systems (IDSs) are significant devices that alleviate threats by detecting malicious actions. In addition, the collaborations between vehicles in VANETs could enhance the accuracy level of detection by interacting with their experiences among nodes. So, distributed ML becomes a highly suitable structure for designing scalable and applicable collaborative detection techniques over VANETs. Therefore, this paper proposes a brain storm optimization with a deep learning-based intrusion detection system (BSODL-IDS) for VANET. In the presented BSODL-IDS technique, a primary stage of BSO based feature selection process is involved in it. For intrusion detection, the BSODL-IDS model exploits the long short-term memory recurrent neural network (LSTM-RNN) model. The Adamax optimizer is utilized at the last stage for the hyperparameter tuning of the LSTM-RNN technique. The experimental validation on the benchmark dataset illustrates the BSODL-IDS method's supremacy over other DL approaches.