Supervisors

Suitable for

Abstract

In response to the increasing centralisation of the World Wide Web and the concerns such centralisation raises (e.g. privacy violations, censorship, loss of individual autonomy, etc), decentralised personal online data stores (pods) have emerged as a promising alternative. A leading example is Solid (https://solidproject.org). It provides a new paradigm to solve issues with privacy-friendly data usage at individual level, but challenges still exist for privacy-preserving computations that involve collective data. We have proposed an architecture targeting exactly this issue, Libertas [1], and have made further exploration in improving the efficiency. That forms a foundation for the problem, but several new interesting open questions emerge. For example: - How to further improve the efficiency of the computation? - How to better facilitate users’ trust selection, such as forming provenance or undeniable records of agents’ behaviour history? - How to allow users to express exercise autonomy, such as by allowing more types of restrictions in agent selection and computation performance? - What technical improvements can lower users’ cognitive burden when setting up and using the architecture?

In this project, the student will explore creative or user-friendly approaches to improve the privacy-preserving collective computation architecture, to address one of the questions above or propose their own questions. The student will not only propose novel solution(s) to address the technical challenges but also empirically analyse the proposed solution and compare with baselines. Interested students are welcome to contact Rui Zhao and Naman Goel to discuss or propose their own ideas related to above (rui.zhao@cs.ox.ac.uk, naman.goel@cs.ox.ac.uk).

Prerequisites: - Willingness to learn new technologies - Good understanding of distributed systems - Good understanding of the Web architecture - Technical experience with at least one of the following will be a big plus: - Solid - Cryptographics in general, or one of the following topics: - Differential Privacy - Multi-Party Computation - Trusted Execution Environment - Familarity with Python or any programming language to work with an MPC framework, such as MP-SPDZ - Familarity with JS/TS, to work with Solid

References:

[1] https://arxiv.org/abs/2309.16365 [2] https://www.w3.org/DesignIssues/PrivateData.html