Artificial Hummingbird Algorithm with Optimal Deep Learning-Based Intrusion Detection on Vehicular Adhoc Networks

Vehicular Ad hoc Networks (VANETs) permit transmission between the vehicle infrastructure, improving Intelligent Transportation Systems (ITS). Vehicles connect wirelessly to transmit the data. However, this transmission is susceptible to numerous attacks, mainly in Vehicle-to-Vehicle (V2V) settings. Intrusion Detection (ID) employing Deep Learning (DL) includes training Neural Networks (NNs) to identify the anomalies and patterns in network traffic, allowing the automatic recognition of potential security attacks and unauthorized events. DL methods, like recurrent NNs (RNNs) and convolutional NNs (CNNs), will efficiently analyze intricate and dynamic datasets for improved ID proficiencies. In this article, an artificial hummingbird algorithm with optimal DL-based ID (AHAODL-ID) technique on VANET is developed. The AHAODL-ID technique exploits feature selection with a hyperparameter selection model for detecting intrusions in the VANET. For data preprocessing, Z-score normalization is employed to scale the input data. Next, the AHA-based feature selection approach is executed for choosing an optimal feature subset. Meanwhile, the Bidirectional Long Short-Term Memory (BiLSTM) approach is implemented to identify various kinds of intrusions. Lastly, the hyperparameter election of the BiLSTM approach involves the design of a Manta Ray Foraging Optimization (MRFO) model. The experimental results of the AHAODL-ID technique are assessed using a benchmark IDS dataset. The obtained values underlined the advanced achievement of the AHAODL-ID technique over other existing models.