Volume 34, Issue 1, 2025
DOI: 10.53555/03276716.2025.11
MLO-Optimized Multi-Scale Fusion Gated Graph Recurrent Network for Securing Clinical IoT Wireless Network and Sensitive Clinical Data
Abstract
In modern healthcare and clinical environments, the rapid integration of connected medical devices and IoT-enabled monitoring systems has significantly increased the vulnerability of clinical networks to cyber threats. The protection of sensitive patient and institutional data in such heterogeneous wireless networks is crucial to ensure both privacy and uninterrupted medical services. To overcome these issues, the proposed work implemented a novel Multi-Scale Fusion Gated Graph Recurrent Network (MS-FGRN) to address anomaly detection. This proposed work includes three major components: (i) a Multi-Scale Temporal Convolution Extractor (MSTCE) for short- and long-term temporal patterns; (ii) a Graph Construction Layer (GCL) used for spatial dependencies; and (iii) a Fusion-Gated Unit (FGU) for adaptive multi-feature integration and heightened anomaly sensitivity. To attain a higher accuracy, the proposed MS-FGRN model’s hyperparameters like filter sizes, graph weights, learning rates, and gate dimensions are tuned using a Modified Lemur Optimiser (MLO). This MLO model is inspired by velocity updates, dynamic step sizing, and leader-follower mechanisms that evaluate the higher optimal solution in hyperparameter tuning. The experimental results are validated using the standard datasets such as NSL-KDD, CIC-IDS2017 and BoT-IoT with classification metrics. The proposed MLO-optimised MS-FGRN attained a higher result in all the metrics than the conventional models. Therefore, the proposed MLO-MS-FGRN established a resilient IDS to secure heterogeneous wireless IoT ecosystems.
Keywords
Internet of Things, intrusion detection system, clinical data, graph neural networks, multi-scale temporal convolution, Modified Lemur Optimiser, Accuracy.