Towards Improving and Integrating Homomorphic Cryptography and Trusted Hardware

In the modern era of outsourced computing, guaranteeing user data security and privacy on data in use is important for ensuring user trust and utilization. Both purely cryptographic and hardware-based solutions have been proposed for protecting user data while it is in use, and each method has its various strengths, weaknesses, tradeoffs, and avenues for future work. Further, the integration of such methods has been explored previously, though issues of scale and trustworthiness remain.

In this dissertation, I present several lines of research around advancing the state of the art in homomorphic cryptography and trusted hardware, both individually and in concert. Specifically, I detail research on hardware acceleration of Homomorphic Encryption (HE), better Private Stream Aggregation and Private Set Intersection protocols, evaluating the efficacy of homomorphic encryption libraries, expanding the limited trusted space of the Intel SGX enclave, and adding integrity protection to hybrid HE/SGX computation. These improvements and innovations promote better guarantees for the privacy of user data.