Detecting CAM Flooding Attacks in Vehicular Networks Using Online K-means Algorithm

Vehicular Networks enable vehicles to establish communication with both other vehicles and fixed road site units for the purpose of sharing critical safety and road traffic information. These networks contribute to the development of a self-aware, efficient, secure, and partially autonomous transportation system by facilitating the exchange and collection of Cooperative Awareness Messages (CAMs). However, vehicular networks are vulnerable to security attacks, where one or more vehicles may flood neighboring vehicles with malicious CAM messages. To effectively detect such attacks, an unsupervised online learning approach, such as our proposed modified version of the online K-means algorithm, can be employed. We evaluate this method using seven datasets, achieving nearly optimal detection performance. Our approach outperforms both the traditional online K-means algorithm and three other clustering methods, yielding higher accuracy and F1-scores.