Deep Learning Approach Enhancing Security in Wireless Sensor Networks
K. Hemalatha, Muhammad Hilmi Amanullah · 2025
Since wireless networks of sensors (WSNs) have so many uses in both the military then the civilian world, they have emerged as a major area of study in computer science. Usually, a WSN is made up of several sensor nodes that gather and send data to a central location. Nevertheless, the distinct features of WSNs-like resource-constrained nodes, diverse deployment approaches, and multi-hop communication-present serious security issues. Maintaining the confidentiality of these networks requires making sure that they are protected from unwanted access. In order to overwhelmed these obstacles and guarantee the dependability and security of WSN operations, an intrusion detection system (IDS) is essential. Even while IDS systems frequently use Machine Learning (ML) techniques, their effectiveness frequently suffers when handling unbalanced attack data. In order to improve detection accuracy then get beyond this restriction, this study suggests an intrusion detection system (IDS) constructed using Deep Neural Networks (DNNs). Using the cross-correlation approach, the most pertinent characteristics are chosen from the dataset. Then, utilizing these chosen attributes, a customized DNN architecture is created to identify different kinds of invasions. Training and testing of the suggested model remain conducted using the WSN-DShacking dataset. The experimental findings show good performance, with 97.29 % accuracy, 98.62 % precision, and 96.67 % recall. In detecting and categorizing network assaults in WSNs, the suggested DNN model performs better than both deep learning and conventional machine learning models.