Randomization effect for enhanced IoT-network intrusion detection systems

The rapid growth of Internet of Things (IoT) devices has introduced significant security challenges, as traditional network intrusion detection systems (NIDS) are often inadequate for IoT environments. This paper presents the effect of randomization on the performance of different machine learning-based IoT Network Intrusion Detection Systems (IoTNIDS) for IoT networks. Combinations of XGBoost, Random Forest, and CNN + LSTM models, both with and without randomization, were trained and tested on relevant IoT datasets to achieve an optimal solution. The findings demonstrate that the proposed machine learning-based NIDS provides a robust and efficient solution for IoT security, capable of handling various types of network intrusions. The performance comparison across datasets revealed that no single model performed optimally for all scenarios, emphasizing the importance of dataset-specific customization in NIDS design. The incorporation of randomization proved particularly valuable in enhancing the generalization of CNN+LSTM for unseen data, aligning with the recent findings in the literature.