Enhancing Cyber Threat Detection on the Internet of Healthcare Thing Devices Using Hybrid CNN-Federated Learning Algorithms
Open Access DepositedThe introduction of the Internet of Things (IoT) technology into how daily routines has sparked a technological revolution across all sectors of society. The inclusion of these IoT technologies in the healthcare sector eventually resulted in the emergence of a new subdomain of the IoT called the Internet of Healthcare Things (IoHT) technologies, which collect and transmit health-related information to healthcare providers in near-real time. While IoHT has the potential to offer benefits by bringing better care and streamlining data management, it can also create new cybersecurity challenges as well as privacy issues. This praxis examines the critical issue of IoHT cybersecurity and privacy. It focuses on building a robust privacy-preserving Intrusion Detection System (IDS) by integrating properties of Convolutional Neural Networks (CNNs) and Federated Learning (FL). In general, most solutions for cyber threat detection in the IoT world, such as signature-based and anomaly detection techniques, cannot deal by themselves with the sophistication and dynamics of modern cyber threats. There is a need for a more sophisticated and adaptive approach solution to ensure that critical IoHT components are secure. Furthermore, specifically in healthcare, it is not just about data integrity of data but also about keeping patient safe and their valuable health information confidential. This praxis aims to improve the security and privacy of IoHT devices by developing a hybrid deep learning model that includes the federated learning method. Federated learning is a machine learning approach that operates in a decentralizing manner by allowing multiple devices to collaborate in training a global model while retaining the data on each device at a localized level. This is generally a robust technique that could be potentially suitable for healthcare environments, where data privacy and security are of utmost importance. This is because it reduces the chance of vital data leakage and fulfills rigid healthcare regulations. The binary and multi-class classification tasks have been tested for the CNN-FL model using the ECU-IoHT dataset. Various metrics for evaluating the model’s performance included accuracy and false alarm rates. For the binary classification, the CNN-FL model achieved its highest accuracy rate at 99.20% and its lowest false alarm rate at 1.42%. For the multi-class classification, the highest accuracy rate achieved was 97.60%, and the lowest false alarm rate achieved was 0.55%. These results underscored the model’s effectiveness at detecting and classifying cyber threats in IoHT devices. Furthermore, the performance of the CNN-FL model was compared to centralized deep learning models (CNN, DNN, and RNN), and the results showed that the federated learning approach held high performance while offering significant privacy advantages.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.