Specific aspects of designing an information security system to protect IoT networks from attacks

Abstract

In today’s rapidly developing Internet of Things (IoT), the issue of ensuring reliable information security is of particular importance. IoT networks are characterized by many interconnected devices that generate significant amounts of traffic and data, which complicates the processes of monitoring and timely detection of cyberattacks. Traditional protection methods are insufficient, since they do not consider the high level of traffic dynamics and the specifics of resource-limited devices. In this context, a promising direction is the application of machine learning methods for designing intelligent attack detection systems that can respond adaptively to new threats. The study focuses on analyzing individual aspects of designing an information security system for IoT networks, particularly the use of the Decision Tree and K-Nearest Neighbor (KNN) algorithms. Both methods are well-known approaches in data classification, demonstrating high efficiency in intrusion detection tasks. The Decision Tree algorithm allows you to create a hierarchical decision-making model that provides a clear interpretation of the results and the ability to explain the classification process. The KNN method, in turn, is based on the analysis of the proximity of objects in a multidimensional feature space, which allows you to adaptively classify new input data, relying on already known examples. Special attention is paid to forming feature sets that affect the system's accuracy. The specifics of IoT traffic are considered: the protocols' heterogeneity, the data transmission frequency variability, and the presence of standard and attack behavior patterns. Representative data sets containing examples of both legitimate requests and modern common attacks on IoT networks were used for experiments. The test results showed that both algorithms provide a satisfactory level of accuracy, but demonstrate different characteristics in the context of speed, noise sensitivity, and computing resource requirements. The scientific novelty of the work lies in the comparative analysis of the effectiveness of the Decision Tree and KNN methods for detecting attacks in IoT networks and in identifying key aspects of designing security systems focused on working in conditions of limited resources. The practical significance lies in the possibility of using the obtained results to create lightweight IDS systems suitable for integration into modern IoT environments. Thus, the proposed approach increases the protection level of IoT networks, ensuring timely detection of potential attacks and reducing the risks of compromising data and services.