An intelligent cyber-attack detection framework for healthcare cyber-physical systems using optimized graph-based deep learning

Abstract

Healthcare Cyber-Physical Systems (HCPS) integrate intelligent medical devices for efficient patient monitoring and treatment. System connectivity and data flow within these networks create greater risks for professional cyber-attackers. Healthcare facilities require modern detection methods beyond static security tools to address contemporary complex evolving security threats in their dynamic systems. Intelligent adaptive detection systems are urgently demanded by healthcare institutions that require precise identification coupled with swift responses and minimal incorrect system alerts. The protection of sensitive patient data alongside HCPS infrastructure reliability represents the core reason for conducting this research. The proposed system uses BitonicX Filtering (BF) during its initial stage for noise-resilient preprocessing of data before employing Gabor Wavelets Hilbert Transform (GWHT) for extracting rich feature sets from physiological signals. Fundamental features received from the second layer are processed by the Clifford Steerable Graph Sample and Aggregate Convolutional Neural Network (CSGSACNN). The learning parameters of the framework reach their best performance levels through optimization with the Human Memory Optimization Algorithm (HMOA). The full pipeline aims to detect threats yet maintain excellent precision through comprehensive feature extraction which supports healthcare system performance at peak levels and increases scalability. The attack detection model operates at a 99.4% accuracy level using a precision of 98.4% along with a recall of 98.0% and a specificity of 97.9%. The framework delivers quick performance along with minimal network usage and reaches superior results in every essential measurement aspect. Healthcare patient security benefits from the CSGSACNN-HMOA framework because it enables smart threat detection which delivers high precision alongside efficient threat processing. The model operates to safeguard data integrity and patient safety at the same time as facilitating secure deployment in scalable healthcare environments.