Fuzzy Clustering for Misbehaviour Detection in VANET

Security in Vehicular Ad Hoc Networks (VANETs) has become a hot research area due to its life-saving characteristic. It mainly depends on the reliability of data exchanged by vehicles (sent, received, forwarded) that includes crucial information such as geographic position, speed, direction, etc. Indeed, several security threats might affect the reception of data itself by dropping, altering or delaying the exchanged messages. In this paper, we present a lightweight misbehavior detection schema for detecting and identifying selective forwarding attackers in VANETs for geographic routing protocol (GPSR). According to this attack, malicious vehicles behave like normal nodes but selectively drop messages. Identifying such attacks is challenging and sometimes impossible. Our designed schema adopts a fuzzy clustering mechanism called Fuzzy-MDS to differentiate malicious vehicles from legitimate vehicles. It initially establishes a preprocessing phase that allows retrieving a set of features from a trace file generated by running the simulation. Based on extracted features, Fuzzy-MDS uses FC-Means algorithm to get clusters of normal and attacker nodes. To enhance the efficiency of our detection schema, a decision threshold is tuned to well identify the misbehaving vehicles. Our preliminary simulation results and analysis show that an accuracy of 87% could be achieved with a low false positive rate under attack's intensity of 25% of the total number of nodes.