A Blockchain-Based Framework for Building and Sharing Machine Learning Models

Abstract

Artificial Intelligence (AI) and blockchain are two disruptive technologies proving to be a powerful combination, improving virtually every industry. Research in Artificial Intelligence has come a long way, with many corporations, organizations and individuals greatly benefiting from Machine Learning services. This research work focuses on creating a Blockchain-based framework for building and sharing Machine Learning models. We break the framework into three components to understand it: Publishing AI Services, Contribution to AI Services, and Fetching Private Predictions from AI Services. In the first component, the framework allows corporations, organizations and individuals to publish their AI Services. In the second component, the framework incorporates machine learning practitioners who can contribute to AI Service Publishers by building models and datasets in exchange for a reward. Thirdly, the framework allows corporations, organizations and individuals with problems that could be solved with AI to solicit solutions from AI Service Publishers without sharing their data. This research utilizes the benefits of existing multi-party computation (MPC) libraries to build Machine Learning models that provide predictions on encrypted data.