A privacy preserving secure predictive framework for iot based health cloud system using herde and msnb

Abstract

In recent days, one of the most popular emerging technologies in the it industry is the internet of things (iot). iot is described as interlinked physical devices that are both connected and smart. sensors are embedded in interconnected physical devices using wired or wireless networks and interact with each other. the key features of iot are interconnectivity of devices, smart, dynamic nature, sensing, enormous scale, heterogeneity, and security. cloud provides many services to a customer over the network such as storage, application, and database. iot offers an extensive range of field applications for continuous monitoring across several domains and health care is one among them. especially, with the advent of the iot-cloud-based devices, iot is established in the field that processes a very high amount of data. the health care system is one of the emerging applications of the iot-cloud. many research works are carried out in ensuring the privacy of the patient data. the main issues in the iot-cloud based health system remain on the security of data along with computation overheads. predicting disease using patient data from the iot device is another demanding aspect of health systems. in this work, a novel homomorphic encryption with random diagonal elliptical curve cryptography integrated with multi-nomial smoothing naive bayes (herde-msnb) is proposed to provide effective security and predict the disease over patient data in the iot health cloud system. the cryptic framework in the proposed architecture involves the encryption and decryption of the patient data along with keywords through the herde algorithm. the medicinal person deciphers the encrypted data and performs the prediction through the msnb model. the uci repository dataset is employed to predict the performance of the security and prediction model. from the analysis, it is observed that the proposed architecture is effective in providing security and disease prediction than the existing models with less processing time, computational cost, and