Abstract
One of the most significant challenges of vehicular ad-hoc networks (VANETs) is establishing reliable connections with the network infrastructure despite the participation concern of attacking vehicles in routing. Vehicles need a proper defense mechanism against all types of attacks, even if a reliable path-planning strategy accompanies them. This paper proposes a reliable score-based routing protocol using a fog-assisted intrusion detection system (RSR-IDS) in VANETs. First, RSR-IDS pre-processes data using minimum-maximum normalization and Pearson's correlation coefficient. The IDS is trained using three machine learning-based algorithms and a voting technique to reduce false detection. These algorithms include the decision tree, random forest, and extra trees. Deploying the IDS in the fog server solves the data diversity problem in the classifier training. Therefore, RSR-IDS detects abnormal data accurately to calculate the untrust score (US). Then, RSR-IDS selects a route with the lowest total USs and hop count compared to others for communications. RSR-IDS is evaluated based on the accuracy, F1-score, false negative rate, packet delivery ratio (PDR), packet loss ratio, end-to-end delay, and throughput criteria using OMNeT + + and the UNSW-NB15 dataset. The significant improvements in RSR-IDS include 14.1% in accuracy, 11.4% in F1-score, and 5.4% in PDR regarding various vehicle densities.