Optimizing Feature Selection for IOT Intrusion Detection Using RFE and PSO

Abstract

Abstract: Internet of things (IoT) and DoS attacks are two of the modern subjects currently being discussed and studied. In this paper, An approach the defense algorithm of IDS for IoT networks’ security development contrary to attacks of DoS applying unusual ML and diagnosis has been presented. An anomaly detection is used in the provided IDS to control network traffic in an ongoing way for deviations from usual profiles. Four observed classifier algorithms have been applied: k-Nearest Neighbor (kNN), Support Vector Machine (SVM), Random Forest (RF), and Decision Tree (DT). Two feature selection mechanisms, which are Particle Swarm Optimization Algorithm (PSO) and Correlation-based Feature Selection Recursive Feature Elimination (RFE) have been used to compare their performances. The dataset of IoTID20 has been used, one of the most currently used to diagnose anomalous tasks in IoT networks, for checking our model. The best results were obtained using RF and kNN classifiers that were trained with features selected by RFE. kNN benefits from the smaller feature space since it focuses on distance measures, which are more successful with a refined set of features. RF improves decision-making by focusing on the most informative features, resulting in better overall performance. RFE notably improved kNN and DT accuracy, while SVM showed consistent results regardless of the feature of selection. These results highlight the importance of feature selection in optimizing classifiers for IoT intrusion detection , and achieved perfect scores (1,00) across all metrics.The aim from this paper is to enhance intrusion detection in iot networks by designing adual stage feature selection method based on RFE and PSO.