In today’s digital landscape, Internet of Things (IoT) networks and systems are increasingly ubiquitous. However, securing and protecting these networks has become a major concern. Addressing this issue necessitates sophisticated techniques for detecting malicious nodes that can autonomously learn and adapt to the IoT environment. In this work, feature selection strategy combined with a Machine Learning (ML) classifier is proposed to detect IoT attacks. It integrates various feature selection models to identify the pertinent attributes from network traffic data, thereby enhancing the accuracy of distinguishing between normal and malicious behaviors. To access the efficacy of the proposed approach, experiments are carried out using a Support Vector Machine (SVM) classifier with different kernel options and assessed its performance to identify the most effective kernel for detecting malicious IoT nodes. The outcomes demonstrate that proposed approach achieved high accuracy and robustness in differentiating between benign and malicious IoT nodes through network traffic analysis. The consistent performance of the SVM classifier across various attack types highlights its versatility and effectiveness in malicious IoT node detection.