An Anomaly Detection Model Using Principal Component Analysis Technique for Medical Wireless Sensor Networks

A Wireless Medical Sensor Network (WMSN) connects autonomous nodes such as (sensors and actuators) existing on the body, or under a person's skin. The network usually extends over the entire human body and nodes are connected via a wireless communication channel. WMSN sense human physiological signs and monitor a patient's health status. The framework and preliminary experiment on developing an anomaly detection model for ubiquitous patient and healthcare monitoring in medical wireless sensor networks (MWSN s)were presented in this study. The architecture is a combination of an improved data mining method and machine learning algorithms using modern fusion methods. Being that MWSNs are highly susceptible to failures due to certain limitations, such as low energy resources, poor reliability, low computational resources, and considerable susceptibility to post-deployment security attacks, the proposed model is an anomaly detection method for MWSNs for the detection of anomalies in an adaptive manner with high accuracy while maintaining resource constraints using two phases. First, an anomaly-based detection model was created using Principal Component Analysis (PCA) technique to reduce dimensionality, prevent overfitting, and increase detection accuracy. Second, the detection accuracy of the proposed model was evaluated and compare before and after PCA integration. The experimental study showed that the proposed model can rapidly identify sensor anomalies with high accuracy.