Decentralized machine learning for secure data sharing

• Main Author: Lim, Jian Cheng
• Other Authors: Mao Kezhi
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Computer science and engineering::Computing methodologies::Artificial intelligence
• Online Access: https://hdl.handle.net/10356/140533
• Institution: Nanyang Technological University

Machine Learning is getting incorporated into all industries nowadays and is simplifying the way everything works. As we move on to a big data era, machine learning becomes commonly used across various sectors. However, in a standard machine learning process, the training data must be gathered from different entities and stored in a single server. With that, only a limited amount of data can be shared due to the sensitive information contained within the data itself. This makes machine learning inefficient as it is unable to learn from those untapped data. Consequently, decentralized machine learning should be utilized to resolve this privacy limitations. This research explores the feasibility of decentralized machine learning through passing of the model’s parameters without putting all the training data together. Experiments making use of neural network was done to inspect the effect of the training parameter of a model.