An Efficient AI and Blockchain Integrated Approach for Healthcare Management

Abstract

Today, many remote patients are ensuring reliable treatments by using smart wearable devices on their bodies. Thus, the effectiveness of the healthcare industry has remarkably grown. However, it is affected by breaches in data security which had been a prime concern owing to the incredible rise in patient numbers. Intruders intercept the data over the channel, and system and tamper with the data which puts the privacy of the medical records at stake. The introduction of the 4.0 medical industry has enhanced the mode of diagnosis and treatment approaches. Medical practitioners or experts access digital information regarding the patient’s condition to administer accurate and fast treatment. However, spoof attacks, manipulation, and hijacking are common vulnerable attacks seen on wearable gadgets. Tampered data passed to the concerned medical expert may put the patient's life at risk. Due to its transparency, security, and being an immutably decentralized network, blockchain is utilized to store information from patient’s wearable devices. The authenticity of the information from wearable gadgets entering the blockchain can be effectively detected using Artificial Intelligence (AI) through machine machine-learned classifier. This paper introduces an AI and blockchain-based highly secured and trustworthy framework to eliminate malicious samples and allow authentic information to flow in the network which can be analyzed by the medical experts and followed by the patients. The dataset samples from the unbalanced WUSTL EHMS 2020 dataset considered for evaluation are efficiently processed by selecting significant features from 43 attributes. The dataset was augmented to balance, extending the lower class (attack) samples. The features were normalized, reduced in dimension and classified using the Support Vector Machine (SVM). Experimentation showed that the malicious samples were distinguished from the normal samples at an accuracy above 98% which is better than other recent competing researches.