Analysis and Evaluation of ToN IoT Windows and Garage Door Datasets

Software security has become increasingly important due to the increasing incidence of system breaches and sophisticated threats, especially in Internet of Things (IoT) systems. The development of effective intrusion detection systems remains critical to ensure privacy, security, and availability in IoT frameworks. This work presents a comprehensive statistical and machine learning-based analysis of the ToN IoT datasets for Windows 7, Windows 10, and Garage Door, a useful resource developed for training and benchmarking the performance of AI-based cybersecurity solutions. The contributions of this work include a complete characterization of the datasets, a performance comparison using different classification algorithms, and a study of anomaly detection methods. The results demonstrate the predictive power of selected correlated features and expose the performance limitations of commonly used anomaly detection algorithms when applied to complex IoT datasets.