NAIBI: A Neighbor-vehicle Approach for Anomaly Detection of Basic Safety Message Falsification in Internet of Vehicles Using Machine Learning

Abstract

The alarming rate of fatality and injuries recorded through road accidents call for the deployment of intelligent transportation system (ITS). The internet of vehicles (IoV), being the backbone of ITS, provides vehicles with standards and protocols to disseminate basic safety messages (BSMs) containing kinematic information to other vehicles and infrastructures, making the IoV a complex network and therefore susceptible to cyberattacks. Despite employing public-key infrastructure (PKI) to ensure BSMs are digitally signed and authenticated, insider attackers can still falsify BSMs and cause chaos in the network. The research community has contributed by proposing data-centric approaches however, the over-reliance on one vehicle BSM data for training and inference gives the attacker an upper hand. To address these drawbacks, we proposed a machine learning-based neighbour-vehicle approach for anomaly detection of BSM falsification in IoV (NAIBI) and demonstrate its superiority over the state-of-the-art which exceeds 99% in accuracy, precision, recall and F1-score.