Privacy-Preserving Learning Models for Communication: A tutorial on Advanced Split Learning

Abstract

Along with the development of the internet, there is an explosion of data that is assembled from diverse sources such as mobile phones or sensors with unprecedented volumes. Big data paves the way for devising state-of-the-art Machine Learning models, which are employed for various tasks such as predicting or analysis, and bring massive impacts on people's daily lives. However, traditional learning architectures pose risks of privacy leakage. In deal with this problem, Split Learning was introduced. Split Learning is an emerging Distributed Collaborative Machine Learning approach with privacy-preserving nature. In this work, we first introduce the fundamentals of Split Learning and then describe major challenges of current learning models. We then depict modern Split Learning and their potential applications in variety fields.