A Hybrid Framework for Intrusion Detection in Healthcare Systems Using Deep Learning

Area of Science:

• Cybersecurity
• Artificial Intelligence
• Healthcare Informatics

Background:

• Increasing network traffic and data volume challenge traditional intrusion detection systems (IDS).
• E-healthcare systems require robust security for patient data accuracy, confidentiality, and integrity.
• Existing AI-based IDS often use outdated data, leading to false positives and frequent obsolescence.

Purpose of the Study:

• To propose ImmuneNet, a hybrid Deep Learning framework for detecting the latest intrusion attacks in healthcare data.
• To develop a lightweight, efficient IDS suitable for medical devices and healthcare systems.

Main Methods:

• Developed ImmuneNet, a hybrid Deep Learning framework with <1 million parameters.
• Employed multiple feature engineering processes, oversampling for class balance, and hyper-parameter optimization.
• Benchmarked performance against other machine learning algorithms on CIC-IDS 2017, 2018, and Bell DNS 2021 datasets.

Main Results:

• ImmuneNet demonstrated superior performance on the CIC Bell DNS 2021 dataset.
• Achieved high accuracy (99.19%), precision (99.22%), recall (99.19%), and ROC-AUC (99.2%).
• Outperformed existing approaches in classifying normal, intrusion, and cyber attack requests.

Conclusions:

• ImmuneNet offers an effective, up-to-date solution for e-healthcare intrusion detection.
• Its lightweight and fast architecture is suitable for IoT deployment in medical devices.
• The framework enhances the security and reliability of patient data against modern cyber threats.