Covid-19 digital contact tracing with sequence embedding

Abstract

This project aims to propose a novel digital contact tracing solution TracingwPrivacy with privacy preserved for users, and deliver a Proof of Concept for the solution. The current contact tracing apps and QR code solutions are either centralised systems that send every users’ data to the government database or partially decentralised systems where the government maintains COVID-19 patients’ and all their contacts’ data. Different from the current solutions, this project proposes a completely decentralised system, where the government has access to embeddings of users’ trajectories, not the original data. TracingwPrivacy addresses the rising concern of data privacy in digital contact tracing. Firstly, existing embedding algorithms were researched. SGT (Sequence Graph Embedding) was found suitable. Then a multi-agent simulation of the mobile system was built since we could not take data from the real world. The simulation generated data, which were embedded by SGT and sent to a server. The server would compute the similarity between the embeddings and return similar ones upon the frontend’s request. Finally, the advantages and limitations of such a system are discussed.