Subsystem of anomaly detection in the Smart House system based on machine learning

With the deepening implementation of Smart Home systems, the role of anomaly detection subsystems becomes increasingly important for ensuring the security and stability of these complex environments. This paper proposes a new anomaly detection subsystem for Smart Home systems, based on advanced machine learning technologies. The architecture of this subsystem is designed to process various data streams generated by Internet of Things (IoT) devices, utilizing packet preprocessing to optimize data before further analysis. The application of the Random Forest algorithm allows for the construction of a machine learning model for effective anomaly detection in the system. To evaluate the effectiveness of the proposed subsystem, the CICIDS2017 dataset is utilized, which is divided into training and validation sets. Comparative analysis is conducted with the J48 tree algorithm in detecting various types of cyberattacks, such as Denial of Service (DoS), Probe, Remote to Local (R2L), and User to Root (U2R). The proposed subsystem aims to enhance the security and reliability of Smart Home systems by facilitating timely detection and response to potentially dangerous anomalies. This work represents a significant contribution to the field of smart systems as it addresses the security issue within the smart home environment, where a large number of connected devices are typically characterized by limited resources and increased requirements for confidentiality and availability. The application of machine learning methods for anomaly detection enables the identification of unusual and potentially hazardous interactions between devices and the network, indicating attacks or security breaches. Particular attention should be paid to the experiment results, which demonstrated the high effectiveness of the proposed system compared to traditional methods. Anomaly detection using the Random Forest algorithm proved to be effective in various attack scenarios, providing high accuracy and a low error rate. This suggests the potential use of such approaches for protecting smart systems in the future.