A Novel Homomorphic Encryption and Consortium Blockchain-Based Hybrid Deep Learning Model for Industrial Internet of Medical Things
A Novel Homomorphic Encryption and Consortium Blockchain-Based Hybrid Deep Learning Model for Industrial Internet of Medical Things
复制标题
DOI：
10.1109/tnse.2023.3285070
复制
发表时间：
2023-09-01
影响因子：
6.6
通讯作者：
Fortino, Giancarlo
中科院分区：
文献类型：
--
作者：
Ali, Aitizaz;Pasha, Muhammad Fermi;Fortino, Giancarlo
由于电子病历（EMR）中包含的数据的价值和重要性，因此电子病历（EMR）的安全性是互联网上最关键的加密应用之一。尽管基于区块链的医疗保健系统可以为EMR提供安全性、隐私性和不变性，但现有方案存在一些突出的安全性和延迟问题。例如，一些研究人员使用区块链作为存储工具，这增加了延迟并对区块链性能产生不利影响，因为它存储了每个交易的副本。分布式账本还需要适当的空间和计算能力，以增加数据大小。此外，现有的基于医疗保健的方法通常依赖于连接到云的集中式服务器，这些服务器容易受到拒绝服务（DoS）、分布式拒绝服务（DDoS）和共谋攻击。本文提出了一种新的基于混合深度学习的同态加密（HE）模型，用于工业医疗物联网（IIoMT），以科普使用联盟区块链的此类挑战。将HE与拟议的IIoMT系统集成是这项工作的重要贡献。在将存储外包给云的同时使用HE提供了对加密的EMR数据执行任何统计和机器学习操作的独特设施，从而提供对共谋和钓鱼攻击的抵抗。我们提出的模型在云中使用了一个预先训练好的混合深度学习模型，并将训练好的模型部署到基于区块链的边缘设备中，以便使用EMR对本地模型进行分类和训练。这进一步取决于与联盟区块链连接的每个边缘和物联网设备的私有数据。所有获得的本地模型被聚合到云端以更新全局模型，该全局模型最终被传播到边缘节点。我们提出的方法提供了比传统模型更多的隐私和安全性，可以为用户提供高效率和低端到端延迟。使用基准性能指标，这表明我们提出的模型提供了增强的安全性，效率和透明度进行比较模拟分析与国家的最先进的方法。
Securing Electronic Medical Records (EMRs) is one of the most critical applications of cryptography over the Internet due to the value and importance of data contained in such EMRs. Although blockchain-based healthcare systems can provide security, privacy, and immutability to EMRs, several outstanding security and latency issues are associated with existing schemes. For example, some researchers have used the blockchain as a storage tool which increases delay and adversely affects the blockchain performance since it stores a copy of each transaction. A distributed ledger also requires appropriate space and computational power with increased data size. In addition, existing healthcare-based approaches usually rely on centralized servers connected to clouds, which are vulnerable to denial of service (DoS), distributed DoS (DDoS), and collusion attacks. This paper proposes a novel hybrid-deep learning-based homomorphic encryption (HE) model for the Industrial Internet of Medical Things (IIoMT) to cope with such challenges using a consortium blockchain. Integrating HE with the proposed IIoMT system is a vital contribution of this work. The use of HE while outsourcing to the cloud the storage provides a unique facility to perform any statistical and machine learning operation on the encrypted EMR data, hence providing resistance to collusion and phishing attacks. Our proposed model uses a pre-trained hybrid deep learning model in the cloud and deploys the trained model into blockchain-based edge devices in order to classify and train local models using EMRs. This is further conditioned on the private data of each edge and IoT device connected with the consortium blockchain. All local models obtained are aggregated to the cloud to update a global model, which is finally disseminated to the edge nodes. Our proposed approach provides more privacy and security than conventional models and can deliver high efficiency and low end-to-end latency for users. Comparative simulation analysis with state-of-the-art approaches is carried out using benchmark performance metrics, which show that our proposed model provides enhanced security, efficiency, and transparency.