Real-Time Monitoring and Anomaly Detection in Hospital IoT Networks Using Machine Learning

The study covers the vital aspect of strong abnormality detection systems on IoT hospital networks that do not have enough resources available. The study evaluated the effectiveness of four different algorithms- namely, QAE, WOA, PSO-DL, and DL-AD-in improving network security. It aimed at enhancing maximum intrusion detection but bearing in mind that healthcare IoT devices possess minimal computing capabilities. Real-Time Internet of Things 2022 (RT-IoT2022) dataset was used in the experiments, emulating cyber-attacks against smart health systems. This showed that QAE is compactly designed with good precision (0.92) and recall rates (0.88) due to the employing of quantization techniques. Competitive performance was obtained by WOA, which mimicked the social behavior in the nature resulting in average precision equal to 0.85 and average recall equal to 0.78. Particularly, the combined method of PSO-DL showed great performance where they had accuracy of 0.94, recall of 0.91 and f1-score of 0.92. DL-AD acted as a benchmark giving good accuracy results, with precision and recall averaging about 0.89 and 0.86, respectively. Compared with related works, the proposed algorithms, particularly QAE and PSO-DL addressed the specific issues associated with healthcare IoT networks as far as accuracy and recall are concerned with much better performance results in most cases. These offer very useful insights into a quickly changing terrain of healthcare IoT intrusion detection which ultimately leads to improved adaptability for future security measures. In this case, study stress on the use of customized approach for resource limitation health orientated IoT networks.