Intrusion Node Detection and Energy-Efficient Centralized Clustering Using a Similarity-Navigated Graph Neural Network in Wireless Sensor Networks

Internet of Things applications include smart grid systems, smart cities, and healthcare wearing wireless sensor networks (WSNs) for telecommunication of important data sensing. Thus as WSNs become more developed, energy efficiency as well as data routing security gets to be significant. The improved model to be proposed for intrusion detection and clustering in the WSNs is called Cat Hunting Clustering-based Similarity-Navigated Graph Neural Network with Genghis Khan Shark Optimizer known by the abbreviation CHC-SNGNNetGKSO. First, contingent detrimental nodes are detected and removed by the Similarity-Navigated Graph Neural Network (SNGNNet) and the Genghis Khan Shark Optimizer (GKSO) for the hyperparameter optimization of the model. Cat Hunting Clustering ensures safe and effective routing at a low energy rate of 0.04 J and, a high throughput rate of 350 kbps that is achieved following a 98 % detection rate with only a 2 % error rate as the model follows the Cat Hunting Clustering protocol. This leaves the model as a probable possibility of choice in WSNs depending on the results which indicate improvement in both security and power consumption.