IMPROVING EFFICIENCY OF SECURE SERVICES WITH HARDWARE-ASSISTED ATTESTED EXECUTION

Abstract

This thesis proposes security primitives that are based on trusted computing, especially commodity hardware primitives such as Intel SGX-enabled processors, to enhance the efficiency of secure online services. Firstly, we propose a secure timed challenge-response protocol that leverages trusted hardware to establish assurance on data residency. Secondly, we codify an approach called Scramble-then-Compute for enabling privacy-preserving computations with trusted computing at scale, and demonstrate its efficiency on various major data management algorithms (e.g., sort, compaction, selection, aggregation, and join). Thirdly, we introduce a framework that provisions a marketplace for secure outsourced computations, leveraging attested enclave execution to protect the confidentiality of clients' inputs, to ensure the integrity of the outsourced computations, and to warrant even-handed exchanges between the clients and compute nodes participating in the marketplace. Finally, we study the use of commodity secure processors in scaling distributed consensus protocols, and introduce optimisations that allow existing blockchain systems to achieve orders of magnitude higher throughput at scale.