Scholarlink

<p>The rapid evolution of the Internet of Medical Things has transformed modern health care by enabling real-time patient monitoring and intelligent data-driven decisions in smart cities and homes. However, the heterogeneity and interconnectivity of Internet of Medical Things devices expose networks to sophisticated cyber threats, demanding efficient intrusion detection systems with low computational overhead. Existing intrusion detection systems approaches in health care, such as basic autoencoders, support vector machines, and random forests, often suffer from high false positives, limited generalization, and poor adaptability in resource-constrained environments. To address these limitations, this study introduces a novel intrusion detection systems framework GOOSE-simple autoencoder which integrates a simple autoencoder with the GOOSE optimization algorithm, a nature-inspired meta-heuristic technique, to enhance both detection accuracy and computational efficiency. The proposed model was evaluated on the ICU-Internet of Medical Things dataset and compared with baseline models including simple autoencoder, denoising autoencoder, PSO-simple autoencoder, and FA-simple autoencoder. The GOOSE-simple autoencoder achieved an accuracy of 99.94%, F1-score of 0.9995, and required only 50 epochs for convergence. In contrast, simple autoencoder required 300 epochs to achieve 99.75% accuracy, indicating a ~ 83% reduction in training time. Furthermore, GOOSE-simple autoencoder showed marginal yet consistent performance improvements: + 0.08% over PSO-autoencoder and + 0.14% over FA-autoencoder in accuracy, and a significant reduction in false positives and false negatives. This research demonstrates the effectiveness of combining nature-inspired optimization with deep learning to create a scalable, adaptive, and computationally efficient intrusion detection systems for Internet of Medical Things environments. The model’s adaptability to the resource-constrained environments of IoMT makes it a valuable contribution to the security frameworks of smart cities and homes. Further, its lightweight design makes it suitable for real-time deployment in resource-limited healthcare settings, contributing to the next generation of intelligent, secure medical cyber-physical systems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

GOOSE-Optimized Deep Autoencoder for Intrusion Detection in IoMT Networks: A Scalable and Efficient Approach for Healthcare Security

  • 期刊论文

  • 发表日期：

    2025年8月2日

查看全文
  • Pooja Puspita Priyadarshani,
  • Janmenjoy Nayak,
  • Pandit Byomakesha Dash

摘要

The rapid evolution of the Internet of Medical Things has transformed modern health care by enabling real-time patient monitoring and intelligent data-driven decisions in smart cities and homes. However, the heterogeneity and interconnectivity of Internet of Medical Things devices expose networks to sophisticated cyber threats, demanding efficient intrusion detection systems with low computational overhead. Existing intrusion detection systems approaches in health care, such as basic autoencoders, support vector machines, and random forests, often suffer from high false positives, limited generalization, and poor adaptability in resource-constrained environments. To address these limitations, this study introduces a novel intrusion detection systems framework GOOSE-simple autoencoder which integrates a simple autoencoder with the GOOSE optimization algorithm, a nature-inspired meta-heuristic technique, to enhance both detection accuracy and computational efficiency. The proposed model was evaluated on the ICU-Internet of Medical Things dataset and compared with baseline models including simple autoencoder, denoising autoencoder, PSO-simple autoencoder, and FA-simple autoencoder. The GOOSE-simple autoencoder achieved an accuracy of 99.94%, F1-score of 0.9995, and required only 50 epochs for convergence. In contrast, simple autoencoder required 300 epochs to achieve 99.75% accuracy, indicating a ~ 83% reduction in training time. Furthermore, GOOSE-simple autoencoder showed marginal yet consistent performance improvements: + 0.08% over PSO-autoencoder and + 0.14% over FA-autoencoder in accuracy, and a significant reduction in false positives and false negatives. This research demonstrates the effectiveness of combining nature-inspired optimization with deep learning to create a scalable, adaptive, and computationally efficient intrusion detection systems for Internet of Medical Things environments. The model’s adaptability to the resource-constrained environments of IoMT makes it a valuable contribution to the security frameworks of smart cities and homes. Further, its lightweight design makes it suitable for real-time deployment in resource-limited healthcare settings, contributing to the next generation of intelligent, secure medical cyber-physical systems.