An Efficient Blockchain-Based Privacy-Preserving Collaborative Filtering Architecture

Abstract

Information overload is a phenomenon of our days due to the unprecedented penetration of information and communication technologies (ICT) in our daily lives. As a result, people often end up with more options than they can process to choose from and therefore may opt for choices which do not fit best to their preferences. To address these issues, recommender systems (RSs) were proposed and have gained a lot of interest from the research community and industry. However, privacy is a big concern in these systems. While decentralized recommenders can protect privacy, they lack the needed efficiency to be widely adopted. In this article, we use blockchain as the backbone of a decentralized RS, managing to equip it with a broad set of features while simultaneously, preserving user's privacy. We introduce a new architecture, based on decentralized locality sensitive hashing classification as well as a set of recommendation methods, according to how data are managed by users. Extensive experimental results illustrate the performance and efficacy of our approach compared with state-of-the-art methods. In addition, a discussion about its benefits and opportunities provides ground for further research.