A Comparison on Feature Selection Methods using Machine Learning Algorithms for improving the Performance Parameters of RPL-BASED IoT Attacks Classification

Internet of Things (IoT) is a collection of connected gadgets or objects or machines which are connected through Internet. Each object is identified uniquely with their identity such as IP address. RPL is a suitable network layer protocol for resource constrained nature of IoT. RPL constructs DODAG with the number of control messages. Because of limited resources in IoT, RPL is prone to various attacks. Various Machine Learning algorithms are considered for classification of RPL based IoT attacks. The dataset used for classification of attacks are combined NSL-KDD and UNSW-NB15 dataset. Since many features are involved due to combined dataset, selection of features is a tedious task. So, different feature selection methods were carried out to find the appropriate features to achieve good accuracy, recall and precision. Various parameters such accuracy, recall and precision are considered for the evaluation of the proposed algorithms.