1. Blockchain Based Data Provenance for Trusted Artificial Intelligence

By John Soldatos, INTRASOFT International, Luxembourg | Angela-Maria Despotopoulou, INTRASOFT International, Luxembourg | Nikos Kefalakis, INTRASOFT International, Luxembourg | Babis Ipektsidis, INTRASOFT International, Luxembourg

Downloaded: 1972 times

Published: 22 Nov 2021

© 2021 John Soldatos | Angela-Maria Despotopoulou | Nikos Kefalakis | Babis Ipektsidis

Abstract

Data reliability is a prerequisite for the development of effective and trusted Artificial Intelligence (AI) systems in industrial environments. Unfortunately, industrial data tend to be unreliable for a variety of reasons (e.g., environmental influence, background noise, and sensor failures). This chapter presents the advantages of blockchain technologies for tracking, tracing, and boosting the reliability of industrial data. It also reviews different blockchain solutions for digital manufacturing, including data provenance and reliability solutions. The chapter ends-up presenting a complete solution for tracing data and metadata of AI algorithms for industrial applications. The solution ensures the use of “sealed” AI algorithms leveraging the properties that render blockchains resilient to tampering. Its main function is to persist the metadata of the algorithms, as well as their outcomes (e.g., prediction and classification outcomes). As such, it facilitates the implementation strategies that secure AI systems in production lines and boost a trusted AI environment in manufacturing applications.