PRIVACY-PRESERVING COMPUTATION IN HETEROGENEOUS ARCHITECTURE WITH MINIMAL TRUST
Abstract
The emerging heterogeneous systems bring new challenges to the security of system design. Having multiple trusted execution environments (TEEs) in the heterogeneous system from different hardware vendors increase the attack surfaces. In this project, we will investigate how to use cryptographic primitives to build secure yet efficient heterogeneous systems without trusting all the hardware vendors. We will first investigate the scenario where trusted TEEs are only available in a subset of processing units in the heterogeneous system. We will develop protection schemes to securely leverage the untrusted processing unit to improve system performance. We will then develop protection schemes for the scenario where none of the hardware vendors are fully trusted. Further, we will explore using the physical property of hardware for protection without trusting the TEEs’ hardware vendors. We will develop security proofs to analyze the security guarantee of the proposed scheme, prototype the system using commercial devices, and evaluate the performance of the resulting system. We will also develop a methodology for co-designing the workload, protection schemes, and system setup to optimize the performance with a given threat model.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 20, 2023
- Source ID
- FA95502210548
Entities
People
- Wenjie Xiong
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- Virginia Tech