Context-aware data-centric misbehaviour detection scheme for vehicular ad hoc networks using sequential analysis of the temporal and spatial correlation of the consistency between the cooperative awareness messages

Abstract

Vehicular ad hoc Networks (VANETs) have emerged mainly to improve road safety, traffic efficiency, and passenger comfort. The performance of most VANET applications and services relies on the availability of accurate and up-to-date mobility-information, through so-called Cooperative Awareness Messages (CAMs), shared by neighbouring vehicles. However, sharing false mobility information can disrupt any potential VANET application. As cryptographic techniques used to protect CAMs in VANETs are expensive, complicated, and vulnerable to internal misbehaviour, security lapses are inevitable. Although several misbehaviour detection solutions have been proposed, those solutions assume that the VANET's context is stationary, which does not hold for VANETs in a real scenario, as the vehicle's context changes continuously. The use of static and predefined security thresholds in highly dynamic and harsh environments is the major drawback of those solutions. To address this issue, a context-aware data-centric misbehaviour detection scheme (CA-DC-MDS) is proposed, using sequential analysis of temporal and spatial correlation of the consistency between neighbouring vehicles' mobility information. The static thresholds have been replaced by a dynamic context reference model that is constructed online and updated in a timely fashion using statistical techniques. Firstly, the Kalman filter algorithm is used to track the mobility information received from neighbouring vehicles. Then, the innovation errors of the Kalman filter are utilized to construct a temporal consistency assessment model for each neighbouring vehicle, using a box-and-whisker plot. After that, the Hampel filter is used to construct a spatial consistency assessment model that represents the current context reference. Similarly, plausibility assessment reference models are built online and updated in a timely fashion using the Hampel filter and by utilizing the consistency assessment reference model of neighbouring information. Finally, a message is classified as suspicious if its consistency and plausibility scores deviate significantly from the context reference model. The proposed context-aware scheme achieved a 73% reduction in the false alarm rate while achieving a 37% improvement in the detection rate. This demonstrates the effectiveness of the proposed context-aware scheme compared with the existing static solutions.