Detecting Distributed Denial of Service (DDoS) in MANET Using Ad Hoc On-Demand Distance Vector (AODV) with Extra Tree Classifier (ETC)

Abstract

This paper concentrate on an option for mitigating distributed denial of service (DDoS) attacks that can stern consequences in mobile ad hoc network (MANET). Discovering a solution to a DDoS attack has gained research focus but challenges exists in performing attack detection with high accuracy along with developing a mechanism in detecting diverse methods to classify DDoS attack activities and also to classify it as an effective measure. The existing methods have numerous difficulties involving detection system performance limits, system scalability and stability, and the capability to develop large volumes of information. This paper concentrates on ETC with randomized search algorithm to detect attacks categorized as flooding, scheduling, black holes and gray holes, using a machine learning (ML) technique as classifier for understanding the behavior of these attacks and trains the better classification method in the MANET data transmitting dataset. The ETC algorithm employs the traditional top-down construction method to construct an ensemble of unpruned decision or regression trees. It separates nodes by selecting cut points thresholds completely at random, which sets it apart from previous tree-based ensemble approaches. When the data transmitted in the AODV, the behavior of node is analyzed and reported in the dataset as target which is trained through ML method. This AODV with ML proposed model can justify the behavior of network in MANET and classify the attack type for the current application. Moreover, the ML method performance has been developed through hyperparameter tuning which can be evaluated through confusion matrix metrics. This AODV with extra tree classifier (ETC) generate improved accuracy as 98.89% using hyperparameter tuning process in determining the safe data transaction in MANET.