Compiler Cryptosystem Co-Design

<p dir="ltr">In a world increasingly concerned with data privacy, the availability of privacy-preserving computation is more important than ever. Unfortunately, the widespread adoption of powerful cryptographic techniques such as Multiparty Computation (MPC) and Fully Homomorphic Encryption (FHE) is limited by the great deal of expertise they require, and their lack of easy programmability. Alleviating this burden of expertise is the subject of much research from both the programming languages and cryptography communities, and is the goal of this dissertation. To this end, we introduce the philosophy of <i>Compiler Cryptosystem Co-Design</i>, which posits that in order to achieve this goal, we need to <i>build compilers that understand the target cryptosystem</i>, and simultaneously <i>develop cryptosystem abstractions tailored to the compiler</i> rather than the programmer.</p><p dir="ltr">We justify the above claim by presenting four increasingly ambitious examples of applying Compiler Cryptosystem Co-Design to problems in the domain of Fully Homomorphic Encryption computations. We conclude by observing that although this dissertation represents an important step towards making privacy preserving computation universally accessible, the ideas contained therein barely scratch the surface of what is possible with compiler cryptosystem co-design; we present two interesting future directions to explore.</p>