CAREER: Heterogeneous Secure Computation Framework for Large-scale Data Processing

User-generated data is collected widely on a large scale due to its huge value. Large-scale data collection has significant privacy concerns due to the growing data breaches. Different kinds of privacy-enhancing technologies (PETs) exist.

Still, they all have various limitations in efficiency and scalability due to the scale and complexity of real-world applications, requiring large-scale and complex data processing. This project aims to develop methodologies and a theoretical foundation for heterogeneous secure computation for large-scale data processing, where different kinds of secure computation technologies are integrated into a seamless pipeline such that one can achieve the efficiency and scalability that cannot be attained by any existing secure computation paradigm alone.

The heterogeneous secure computation framework resulting from this project will enhance cybersecurity, individual privacy, and national security by significantly improving the efficiency and scalability of secure computations performed on encrypted data. Moreover, this research will support the development of a diverse cohort of Ph.D., undergraduate, and high school students via the novel Research Experience for High Schoolers (REHS) program to be developed from the project and other education and outreach activities.

Underlying the proposed work is a novel idea of integrating the two seemingly disjoint secure computation paradigms: homomorphic computations and confidential computing. The project goal is to, by combining different secure computation primitives into a pipeline of heterogeneous secure computation, achieve the efficiency, scalability, and generalizability that cannot be attained by any individual secure computation primitive.

Three interconnected research thrusts will be investigated for this goal. Firstly, confidential computing and homomorphic computations will be integrated into a pipeline such that one large-scale data processing workflow can be decomposed into several sub-computations that are handled by either confidential or homomorphic computations. Secondly, a heterogeneous trusted execution environment (TEE) will be researched to overcome the security and scalability limitations of existing TEEs, with novel zero-knowledge enclave verification and end-to-end attestation across different TEEs.

Finally, parallel and distributed computing will be studied in the context of the new heterogeneous secure computation such that the unique challenges of applying parallel and distributed computing in this new heterogeneous framework are addressed. The research outcomes will be an end-to-end heterogeneous secure computation framework with high efficiency and scalability that are comparable to the plain computations without any security or privacy guarantees.

The project outcomes will include (1) open-source projects implementing the framework, (2) scientific articles presented/published at conferences and journals, and (3) novel education materials for REHS and a new graduate-level course in homomorphic and confidential computing.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.