Hierarchical clustering using a Kohonen neural network for secured wireless sensor networks

Abstract

In today's world, wireless sensor networks (WSNs) have found widespread application in environmental monitoring systems, smart cities, healthcare, and military technologies. However, despite their flexibility and scalability, WSNs remain vulnerable to a wide range of cyber threats, namely, from data interception to routing attacks and node spoofing. This study explores an approach to information protection in WSNs based on node clustering using the Kohonen neural network.

Modeling of clustering in wireless sensor networks using the Kohonen neural network has demonstrated significant effectiveness in solving the problem of adaptive network structure management. The generated topological maps show the network’s ability to self-organize, form stable clusters, and integrate new sensors without disrupting classification logic. The Kohonen network successfully distributes sensors based on behavioral features, creating stable and adaptive clusters. The introduction of a new sensor and its automatic assignment to a cluster confirms the model capacity for dynamic response. Cluster formation hat considers residual energy and spatial centralization contributes to energy conservation and extends the network lifecycle.

The proposed clustering approach based on the Kohonen neural network enables an efficient, flexible, and adaptive WSN structure, especially under dynamic operating conditions. Automatic determination of cluster membership for new sensors facilitates rapid response to topology changes without the need for centralized processing. The self-organizing nature of the Kohonen network makes it a promising tool for early detection of cyber threats through node behavior classification. Simulation results demonstrate the advantages of neural approaches in tasks related to information protection, network adaptation, and energy consumption optimization.