Homomorphic encryption for privacy-friendly augmented democracy

Details

Supervisors

Pereira, Olivier ; Méaux, Pierrick

Faculty

Ecole polytechnique de Louvain

Degree label

Master [120] : ingénieur civil en informatique, à finalité spécialisée

Abstract

"We could have a senate with as many senators as we have citizens." We are not in early Athens, but well and truly in the 21st century, and even more. Augmented Democracy is a proposal to assist citizens in their democratic duty through a digital twin. People do not have the time, interest or expertise to be experts in every field. Open-data government portals are barely used and initiatives gathering citizens' opinions reach only limited levels of participation. Following these two findings, Artificial Intelligence could be used to take over some of the tasks citizens are overloaded with. This is a broad topic that raises issues in many different areas. From an engineering perspective, we focus our study on the discussion whether or not privacy-preserving constructions of such avatars are possible. More technically, we formulate the problem as a Collaborative Filtering recommendation system and solve it with Matrix Factorisation. Finally, we use Homomorphic Encryption and propose two privacy-preserving protocols to address the cold-start problems.