An AI-Driven Cybersecurity Framework for the Internet of Medical Things: A Hybrid LSTM-CNN-GAN Approach

Abstract

The Internet of Medical Things (IoMT) is transforming healthcare by connecting medical devices for realtime patient monitoring. However, the security of these devices remains a significant concern. This paper presents a hybrid deep learning model that combines Long Short-Term Memory (LSTM), Convolutional Neural Networks (CNN), and Generative Adversarial Networks (GAN) to enhance the security of IoMT systems. The proposed model ensures data integrity, detects anomalies, and protects against cyber threats in healthcare environments. Data for this study is collected from various IoMT devices, including wearable health trackers, medical sensors, and smart hospital equipment. It includes timeseries data (ECG, blood pressure, etc.), network traffic data, and medical images (X-rays, MRI scans). Cyberattack data, such as malware logs and denial-of-service attacks, is also incorporated to train the model. Preprocessing steps include normalization, imputation, and labeling to differentiate normal and attack scenarios. The hybrid model uses LSTM for timeseries anomaly detection, capturing irregularities in real-time health data, CNN for medical image security and malware detection, and GAN to generate synthetic attack data for detecting zero-day threats. This multi-layered approach ensures timely detection of health emergencies and cyberattacks, safeguarding both patient data and IoMT devices. Model training uses supervised and unsupervised learning techniques, with performance evaluated using metrics such as accuracy, precision, recall, and ROC-AUC. The hybrid model's effectiveness is validated by comparing it against existing IoMT security methods, demonstrating improvements in detection accuracy and robustness. This approach has critical applications in healthcare, providing real-time threat detection, data protection, and anomaly detection to secure IoMT devices, ensuring patient safety, and minimizing the risk of cyberattacks.

Ähnliche Arbeiten