Please note: This PhD seminar will take place online.
Miti Mazmudar, PhD candidate
David R. Cheriton School of Computer Science
Supervisor: Professor Ian Goldberg
Peer-to-peer distributed hash tables (DHTs) rely on volunteers to contribute their computational resources, such as disk space and bandwidth. In order to incentivize these node operators of privacy-preserving DHTs, it is important to prevent exposing them to the data that is stored on the DHT and/or queried for. Vasserman et al.’s censorship-resistant publishing system, CROPS, is aimed at providing plausible deniability to server nodes by encrypting stored content. However, node operators are still exposed to the contents of queries.
In this talk, I will present our architecture that uses information-theoretic private information retrieval to efficiently render a server node incapable of determining what content was retrieved in a given request by a user. Our architecture can be integrated with censorship-resistant publishing systems such as CROPS or distributed file systems such as IPFS. Finally, I will discuss the results of our system simulation, including the communication and performance overhead of our system over other systems without this privacy guarantee, as well as with respect to the closest related work.