The interest of tech companies for user's personal data is higher than ever since one of the keys for larger profits are highly personalized services. Starting from 2017, data overpassed oil and became the most valuable resource on earth as five of the world’s most valuable companies now deal in this new black gold's market.

 

However, individual users are also becoming more and more aware about the value of their digital footprints and push for regulations that protect their rights in this highly profitable market (e.g. the EU's GDPR or US' CCPA). Under the increasing pressure of data protection legislation and public opinion, the large industrial players started to invest in alternative approaches which allow extracting aggregated information about user preferences while also providing privacy guarantees. One such alternative approach which is becoming increasingly popular in recent years is Federated Learning (FL). Nevertheless, the current FL architecture has some limitations, namely security and energy-efficiency issues. The goal of this project is to address these two limitations for the widespread adoption of Federated Learning.