Routing anomaly detection in OLSR-based MANETs

Optimised link state routing (OLSR) protocol as one of the four standard routing protocols provided for mobile ad hoc networks (MANETs) is vulnerable to attacks launched by authorised nodes. An anomaly detection system (ADS) that uses a small set of features, is unable to detect different types of attacks. In this paper, we define a set of features for OLSR behaviour to learn all behavioural aspects of this protocol. Furthermore, we propose a Conceptual Data Collection based ADS (CDC-ADS) in which ensemble methods are used to enhance the accuracy of anomaly detection. Data are collected based on four aspects of OLSR behaviour, then an expert model is learned for each aspect. A selection-based aggregation mechanism is used to conclude from votes of the learned models. The experiments show that creating the experts and combining their predictions increase the accuracy of detecting attacks. Also, testing CDC-ADS with various time slot lengths and network speeds shows its robustness.