HySec-Flow: Privacy-Preserving Hybrid Computing Model

Speaker: Judy Fox, Associate Professor and the Program Director of the Ph.D. Program at the UVA School of Data Science

Abstract: Security and privacy issues have received increasing attention in big-data analytics performed on public or commercial clouds as the data often contains identifiable information concerning human individuals. Collaborative team research on large cohorts of patients across multiple institutions is often impeded by the privacy concerns of sharing personal genomic and other health data. Homomorphic encryption (HE) allows users to perform computation directly on encrypted data. However, HE introduces several magnitudes of computational overheads and is not practical today. There is a promising alternative with hardware supporting a trusted execution environment (TEE), in which sensitive data are kept in a secure storage and processed in an isolated environment, called an enclave. Intel's SGX delivers this idea with their latest Ice Lake Xeon Scalable processors offering up to 512 GB enclaves (e.g. Xeon Platinum 8380). We note that many problems do not need every step of the processing to be run in privacy preserving mode with its overheads and restricted hosting. We describe and demonstrate a new approach HySec-Flow that can efficiently mix conventional and SGX-based tasks. We present performance results on a genomic example BWA on older Intel systems and look forward to this and Global Pervasive Computational Epidemiology (GPCE) applications on the new Intel systems. 

Join the event here.