Identification of compromised nodes in MANETs using machine learning technique

Mobile ad hoc networks (MANETs) are faced with various security challenges emanating from malicious attacks. Their dynamic nature make nodes more vulnerable to attacks from either malicious nodes or intruders since there is no fixed infrastructure resulting in each node acting as router to transmit data. Currently, several solutions have been proposed and implemented in different ways aimed at eliminating or reducing these malicious attacks. However, the attacks still persist. Therefore, this paper proposes an efficient security mechanism based on machine learning as a solution that detects and identifies malicious attacks in real-time basis by classifying packets data as either normal or abnormal. To achieve this, we conducted experiments using logistic regression (LR) and a support vector machine (SVM) to choose the best predictive model utilizing the Iris data set. The results obtained show that LR performed better than SVM with an accuracy of 100% detection rate. Thus, LR is better suited for the identification of malicious attacks in MANETs. Furthermore, we proposed and designed a framework to detect malicious attacks in real-time in MANETs based on packet behavior using the LR model and the components were presented. We believe that, if this framework is implemented in MANETs, it could go a long way to reduce the rate of attacks in the infrastructure less network.