Development of Intrusion Detection Models for IoT Networks Utilizing CICIoT2023 Dataset
Nadia Thereza, Kalamullah Ramli
Abstract
The Internet of Things (IoT) is a rapidly growing technology that enables devices to communicate and exchange data with minimal human intervention. However, this growth increases the volume of sensitive data, making it more vulnerable to security attacks. DDoS is a perilous form of attack that targets IoT networks frequently. Intrusion detection systems (IDSs) are a solution for protecting IoT devices by monitoring network activities and detecting real-time threats and attacks. However, implementing IDS in IoT networks presents several challenges, including power and memory constraints imposed on IoT devices and implementation datasets requiring greater comprehensiveness to accurately define the features of IoT networks. Thus, this study developed intrusion detection models using lightweight ML algorithms, such as decision tree, k-nearest neighbors, random forest, and Naïve Bayes, to identify network DDoS attacks. The latest dataset, CICIoT2023, which includes multiple attacks unavailable in previous IoT datasets, was utilized. We evaluated the model’s performances using accuracy, false positive rate, F1-score, recall, precision, and training and testing time usage. The results show that the random forest and decision tree models outperformed other detection models with 100% accuracy. Regarding time usage, the decision tree model outperformed other models, which could classify 2,926,588 instances in 1 second.