Server-Aided Secure Computation With Off-Line Parties

Online social networks (OSNs) allow users to jointly compute on each other's data (e.g., profiles, geo-locations, etc.). Privacy issues naturally arise in this setting due to the sensitive nature of the exchanged information. Ideally, nothing about a user's data should be revealed to the OSN provider or non-friends, and even her friends should only learn the output of a specific computation. A natural approach for achieving these strong privacy guarantees is via secure multi-party computation (MPC). However, existing MPC-based approaches do not capture two key properties of OSN setting: Users does not need to be online while their friends query the OSN server on their data; and, once uploaded, user's data can be repeatedly queried by the server on behalf of user's friends. In this work, we present two concrete MPC constructions that achieve these properties. The first is an adaptation of garbled circuits that converts inputs under different keys to ones under the same key, and the second is based on 2-party mixed protocols and involves a novel 2-party re-encryption module. Using state- of-the-art cryptographic tools, we provide a proof-of-concept implementation of our schemes for two concrete use cases, overall validating their efficiency and efficacy in protecting privacy in OSNs.