Practical Privacy-Preserving Federated Learning

Abstract: Federated Learning (FL) is a popular solution to distributedly train a model on user devices improving user privacy and system scalability. However, 3rd party developers are still lacking practical systems that would enable easy, collaborative, cross-device, and cross-app FL training in the context of mobile environments.

In this talk, I will present our attempt to bridge this gap with a novel end-to-end system that enables intra- and inter-app training on mobile devices with different types of IID and NonIID data distributions, in a secure and easy to deploy fashion. Moreover, I will talk about how Trusted Execution Environments (TEEs) could be utilized on clients for local training, and on servers for secure aggregation, so that model/gradient updates are protected from adversaries.

Meet the Speaker

Dr Kleomenis Katevas is a Machine Learning Researcher and lead of the ML research team at Brave Software where he is focused on designing and building privacy-preserving, ML-based systems. His research interests lie in the areas of Privacy-Preserving Machine Learning, Mobile Systems, and Efficient On-Device ML. He holds a PhD and an MSc from Queen Mary University of London, UK. In the past he has worked as a Research Scientist at Telefonica Research, Barcelona, a Research Associate in the Dyson School of Design Engineering at Imperial College London, UK, a Research Assistant at Queen Mary University of London, UK, and an R&D Engineer at Philips Healthcare (formerly Philips Medical Systems) in Germany. He was also a Co-founder and CTO of Hiver, a London Business School (LBS) startup that analysed the human interactions in social events using mobile sensing technology.

Venue Accessibility

For those unable to use the main entrance steps, there is a ramp leading directly from Russell Square under the main entrance, with doorbell button. This entrance brings users into the lower ground floor foyer. There are lifts in the lower ground foyer to the rest of the building