Open Collections

Kamal, Kasra

Edge computing is a new computing paradigm that performs computations at the edge of the network and is now being used for GPU-accelerated applications due to real-time application constraints. Securing GPU-accelerated applications running on edge devices is necessary to obtain data confidentiality and integrity. However, edge devices are vulnerable to various threats (e.g, malware) due to being distributed and having a large attack surface. The key primitive for protecting applications running on an untrusted platform stack are Trusted Execution Environments (TEEs). Existing TEEs however have focused on securing CPUs without securing GPUs. We design a TEE for ARMv8-based devices with TrustZone security extensions and integrated GPUs. The TEE provides an isolation environment built in Normal World (NW) and provides secure request-level GPU access. We utilize the TrustZone provided TrustZone Address Space Controller (TZASC) to securely partition memory regions. We demonstrate the applicability of our design by evaluating the overheads present using ARM Fast Models.

Text

University of British Columbia. CPSC 449

Vancouver : University of British Columbia Library

10.14288/1.0413167

Science, Faculty of; Computer Science, Department of

Undergraduate

DSpace