Work

Machine Learning-driven Decentralized Infrastructure for Secure and Reliable Clinical Information Management

Year: 2026
Type: article
Abstract: Connected medical devices and wearable sensors have expanded the possibilities for diagnosis, treatment, and continuous patient monitoring. At the same time, they introduce cyber security risks, including Distributed Denial of Service (DDoS), data injection, and man-in-the-middle attacks. Blockchain can support trusted healthcare-data exchange and record keeping, but computational overhead and limited scalability remain important concerns. Integrating Artificial Intelligence (AI) with blockchain offers a way to analyze medical information and improve protection against cyber threats. Machine Learning (ML) and Deep Learning (DL) methods can identify suspicious activity; however, their effectiveness may decline when datasets contain redundant or uninformative attributes. Conventional Proof of Stake (PoS) systems can also face decentralization and scalability concerns when validator selection is not adequately balanced. This work presents FSAIEB-Chain, a secure medical-IoT data-sharing framework combining feature selection, AI, and an enhanced blockchain. Neutrosophic entropy is used to identify informative attributes in the e-health dataset, after which a hybrid Long Short-Term Memory and Deep Neural Network (LSTM-DNN) model categorizes records as malicious or non-malicious. The proposed Stabilized Pure Proof-of-Stake (SPPoS) mechanism then validates records classified as non-malicious to reduce the risk of data-integrity attacks. Once validation succeeds, legitimate health information is placed in the Interplanetary File System (IPFS), reducing the amount of data held directly on-chain and supporting response time and scalability. Simulation findings indicate that FSAIEB-Chain performs favourably against the compared AI-assisted blockchain approaches for secure healthcare-data transfer.
Language: English
Open Access status: green
OpenAlex ID: W7235213347
FWCI:
Cites:
Cited by:
Related to: 10