This paper addresses the crucial challenge of securing interconnected systems amid the widespread integration of Internet of Things (IoT) devices across diverse domains. To counter escalating cyber threats against IoT devices, Intrusion Detection Systems (IDS) have been implemented. The proposed defense system employs machine learning techniques to dynamically enhance multi-class intrusion detection in IoT environments. The study evaluates five distinct machine learning Algorithms, Logistic Regression, K-Nearest Neighbors, Decision Tree, Random Forest, and XGBoost, within the context of multi-class IoT intrusion detection systems. The model, trained on a comprehensive dataset encompassing various intrusion scenarios, demonstrates its ability to detect a wide range of security threats. The research employs diverse evaluation metrics, including accuracy, precision, recall, F1 score, and Cohen’s Kappa, contributing to enhancing IoT security and providing valuable insights into machine learning algorithm efficacy for multi-class intrusion detection. In experimental evaluations, the proposed machine learning system outperforms traditional intrusion detection methods, achieving high accuracy in identifying and classifying multi-class intrusion scenarios. The findings present a robust and adaptive defense mechanism, contributing to the advancement of IoT security, and the insights gained pave the way for future developments in intelligent and self-learning security systems, promoting a more resilient IoT infrastructure and guiding informed decision-making in deploying intrusion detection systems.