Detail

Cryptographic protocols for privacy-preserving data pipelines

In this talk, I will examine the broad problem of how to build pipelines for collecting, storing, aggregating, and publishing data, without sacrificing the privacy of those users that may appear in such datasets. Specifically, we aim to build cryptographic solutions that are high-performance — even for large numbers of clients — and simple to implement — so that they can be understood and used by those that are non-experts in the fields of cryptography. We will discuss two recent works that construct solutions for different stages of the pipeline: the STAR protocol, that allows collecting and aggregating K-anonymous data points from Internet clients; and the FrodoPIR protocol, that allows clients to retrieve items from a server database without revealing the items that they are querying. When compared with previous solutions, both protocols emphasise higher performance, lower running costs, and greater simplicity of design. In addition, both protocols are being actively developed and used within the Brave Internet browser (https://brave.com/), and STAR is being considered for standardisation with the IETF.