A Policy-Agnostic Language for Oblivious Computation
Primary Investigator:
Benjamin Delaware
Qianchuan Ye, Benjamin Delaware
Abstract
Secure multiparty computation (MPC) techniques allow multiple parties to collaboratively compute functions over sensitive data in a privacy-preserving manner. MPC protocols use powerful cryptographic techniques to achieve these privacy guarantees, making them challenging for non-experts to directly use. To address this challenge, several high-level languages have been proposed to make writing such applications accessible. These languages typically require the programmers to embed their privacy policies into the application logic, making it hard to audit the policies, or experiment with different policies.
This poster presents our ongoing development of a privacy-preserving language, Taype, that decouples privacy and functionality concerns. Two key ingredients of this language are oblivious algebraic data types and tape semantics. Oblivious algebraic data types are a form of dependent types with oblivious constructs, that can be used to modularly encode complex privacy policies for structured data. Tape semantics then enforce these policies during execution, enabling applications to modularly compose policies and programs written in a conventional way without compromising privacy.