A Secure Recommendation System for Providing Context-Aware Physical Activity Classification for Users

Abstract

Advances in Wireless Body Area Networks, where embedded accelerometers, gyroscopes, and other sensors empower users to track real-time health data continuously, have made it easier for users to follow a healthier lifestyle. Various other apps have been intended to choose suitable physical exercise, depending on the current healthcare environment. A Mobile Application (Mobile App) based recommendation system is a technology that allows users to select an apt activity that might suit their preferences. However, most of the current applications require constant input from end-users and struggle to include those who have hectic schedules or are not dedicated and self-motivated. This research introduces a methodology that uses a “Selective Cluster Cube” recommender system to intelligently monitor and classify user behavior by collecting accelerometer data and synchronizing with its calendar. We suggest customized daily workouts based on historical user and related user habits, interests, physical status, and accessibility. Simultaneously, the exposure of customer requirements to the server is also a significant concern. Developing privacy-preserving protocols with basic cryptographic techniques (e.g., protected multi-party computing or HE) is a standard solution to address privacy issues, but in combination with state-of-the-art advising frameworks, it frequently provides far-reaching solutions. This paper proposes a novel framework, a Privacy Protected Recommendation System (PRIPRO), that employs HE for securing private user data. The PRIPRO model is compared for accuracy and robustness using standard evaluation parameters against three datasets.