Enhanced Human Activity Recognition based on Activity Tracker Data Using Secure Homomorphic Encryption Techniques

Abstract

The fitness tracker data for human activity recognition is playing an important role in improving health and monitoring the daily movements of human such as health care and fitness. This detects human activities such as step count, sleeps rate, calories burned, and person mood for the data collected by the activity tracker device and mainly helps in identifying the Apnea-Hypopnea Index (AHI) which is caused due to Obstructive Sleep Apnea. Medical data is very susceptible for many external threats that affect its security and privacy. The privacy and security of the data shared across the medical field is very difficult to maintain. To overcome these problems, Homomorphic Encryption techniques are proposed, which is used for privacy preserving out-sourced storage and computation. The encrypted technique helps in protecting sensitive data from outsourced. The two Homomorphic techniques are Gorti’s Enhanced Encryption Scheme and Carmichael’s Encryption Scheme that are used to encrypt the fitness tracker data. The computation is made with the encrypted data without revealing the original data to providers. The analysis is made on the encrypted data by computing the AHI to predict the disease based on the range given in AHI. Thus, security of the data is maintained. The performance of the schemes is measured based on the time taken for encryption and decryption of the medical data and analysis performed on the encrypted data.