Predicting model for identifying the malicious activity of nodes in MANETs

For many applications based on Mobile Ad Hoc Networks (MANETs), the position of the nodes is generally hard to be determined. In sensor networks, for instance, such information may be critical for the MANETs. Additionally, one problem to be faced in this scenario is the fake parameters broadcasted by misbehaving/malicious nodes, which can either compromise results about positioning, or deplete power resources of mobile devices. Therefore, in this paper we propose a model for (1) identifying the fake parameters broadcasted in the network, and for (2) detecting the malicious/misbehaving nodes. The Linear Regression and Variance Analysis (LRVA) are both the basis for the multi-step-ahead predictions in this paper. Through NS-2 and Avrora, we simulated the movement and energy consumption of nodes in a MANET, analyzing the time series of beacon-packets exchanged in the network. As a result of the LRVA employment, the fake parameters broadcasted in the network were detected, with the malicious/misbehaving nodes identified. The simulations presented in this paper show low power consumption, which allows the jointly employment of LRVA with other security techniques in the MANETs.