Detection of Position Falsification Attacks in VANETs Applying Trust Model and Machine Learning

Vehicular ad hoc networks (VANETs) are relatively new networks that focus on intelligent transportation systems (ITS). The interest in this kind of networks lies in the promising challenge to enhance security in vehicular transportation systems trying to alleviate driving problems. However, this technology has many concerns before its implementation, especially in topics related to privacy, network overhead and security. Some approaches have been studied to ensure security inside vehicular networks and protect them from attackers, either external or internal. Among several options, trust models have acquired great importance and good results when detecting misbehaving in the nodes. The present work aims to evaluate parameters used for the computation of trust metrics applying machine learning techniques. Results show the superior discriminative power of the receiver power coherency metric when detecting misbehaving nodes based on fake position attacks. Simulation results show the effectiveness of our proposal in terms of ability to correctly classify well behaved and misbehaved vehicles.