In many applications, analytics and machine learning rely on collected personal data. Recent research in decentralised machine learning, has focused on leaving personal data on user devices, and transmitting only models and updates.
In either approach, proper evaluation of a user's contrbutions is a challenge, as is ensuring rigorous privacy of the user's data. These issues lead us to the question: how can we ensure that distributed learning and data collection are private, while at the same time estimate the contributions made by personal data?
This project will investgate decentralisation and privacy aspects of machine learning and data sharing. What kind of statistical privacy can we achieve in decentralised machine learning? Can we estimate a user's contribution to a decentralised algorithm? Are such algorithms robust to adversarial behaviour? Can blockchain based approaches help to develop democratic, collaborative, but secure machine learning platforms?
Thus, the research has broad scope. Consider applying or getting in touch if you have interest/experience in one or more of the following areas:
Deputy Director, Laboratory for Foundations of Computer Science
School of Informatics, University of Edinburgh
http://homepages.inf.ed.ac.uk/rsarkar/
Please put "FAIRCROWD position" in the title of your email.
The FAIRCROWD project is part of the REPHRAIN National Research Centre on Privacy, Harm Reduction and Adversarial Influence Online