Detection of Minority Attacks in IOT Devices Using Recursive Feature Elimination Technique

Effective intrusion detection systems are in high demand to protect the security and privacy of Internet of Things (IoT) devices and the networks to which they connect. In this study, we use the CICIDS2018 dataset, a widely-used benchmark for network intrusion detection in IoT environments, to investigate anomaly detection in IoT gadgets. Our objective is to identify and classify minority assault data using machine learning techniques. To do this, we employ four widely used classifiers: a Decision Tree, a Random Forest, a K-Nearest Neighbours (KNN) algorithm, and a Support Vector Machine (SVM) technique. These classifiers are trained with data from the CICIDS2018 dataset, which describes the various forms of network traffic produced by IoT devices. Recursive Feature Elimination (RFE) is used to select the most relevant features and improve the models' classification performance. Furthermore, we prevent attacks against the minority group from being noticed by randomly undersampling the sample. This allows us to concentrate on accurately detecting and categorising the very few but essential attacks that represent the most risk to IoT systems. We run tests and analysis to compare the four classifiers' performance in terms of accuracy, precision, recall, and F1- score. The effects of feature selection and data balancing on classification precision are also explored.