Publications

In today's data-driven world, recommendation systems personalize user experiences across industries but rely on sensitive data, raising privacy concerns. Fully homomorphic encryption (FHE) can secure these systems, but a significant challenge in applying FHE to recommendation systems is efficiently handling the inherently large and sparse user-item rating matrices. FHE operations are computationally intensive, and naively processing various sparse matrices in recommendation systems would be prohibitively expensive. Additionally, the communication overhead between parties remains a critical concern in encrypted domains. We propose a novel approach combining Compressed Sparse Row (CSR) representation with FHE-based matrix factorization that efficiently handles matrix sparsity in the encrypted domain while minimizing communication costs. Our experimental results demonstrate high recommendation accuracy with encrypted data while achieving the lowest communication costs, effectively preserving user privacy.

Human Activity-based studies have become an omnipresent research topic in Machine Learning. Considering the countless impacts of human activity on persons' everyday life, we have analyzed the correlation between human activity and their product preferences in our study and proposed that daily human activity could be a metric for product recommendation models. To address this previously unaccounted phenomenon, a new approach is pre sented in our study that gives real-time recommendations to users by observing their activeness in daily life. However, product recommendation systems mostly believe in ratings, and the purchase behavior of users instead of investigating the precious insights of users' daily activities. But we examined smartphones' GPS sensor data using machine learning algorithms to urge insights from users' daily activeness and proposed a model for predicting the product of interest of the purchasers, based on the activeness of their daily life. Moreover, based on our model, we have introduced a prototype of a real-time recommendation system, especially for the retail shops that rely on users' implicit data from smartphone sensors to form product recommendations. For conducting our study, we devel oped an android application that-collects embedded smartphone sensor data and can detect objects to provide product recommendations and product details. Experiment shows, that our proffered daily activeness-based recommendation system using smartphone sensor data, performs with a precision of 66%, but it is also a promising performance because it does not use customers' explicit feedback.

The proliferation of smartphones has become a ubiquitous platform for acquiring and analyzing data. Smartphones' embedded sensors have become an effective source for human spatial and activity-based analysis. Machine Learning (ML) has made significant progress in learning features from these raw sensor data with high accuracy. However, domain experts, knowing ML, can apply machine learning techniques for various aspects. In this research, we have introduced-a smartphone sensor data collection and analysis platform for people in general who have little or no knowledge of machine learning but can avail the services of machine learning for their purpose. We have built an Android application for collecting sensor data and developed an Automated Machine Learning (AutoML) based web platform for data pre-processing, visualization, and analysis. Spatial analysis has been conducted on our AutoML based web application on GPS sensor data. We evaluated the most visited places of our app users using clustering techniques. The experiment shows that the DBSCAN clustering algorithm gives superior performance over K means clustering for our spatial analysis on GPS sensor data.