Dynamic event classification for intrusion and false alarm detection in vehicular ad hoc networks

摘要

Several classification techniques have been proposed as a basis to build intrusion detection systems for vehicular ad-hoc networks. In this paper, we proposed a dynamic event classification technique to categorise communication messages for the purpose of detecting intrusions and false alarms. The contributions of this research are to: 1) propose an efficient binary classification technique to evaluate the plausibility of communication messages in VANETs based on a set of semantic patterns of actions; 2) apply a mechanism to construct association rules that handle the representation of ad-hoc conditions. The proposed technique relies on defining the classification task as an optimisation problem that maximises true-positives and minimises false-positives. A set of experiments have been performed in order to evaluate the proposed technique using two different datasets. The results indicated that our proposed technique outperformed state-of-the-art classification techniques and efficiently detect intrusions and false alarms.