Research for secure on-chip isolated execution environment on commodity graphics processing units

Abstract

In the last decade, the dedicated graphics processing unit (GPU) has emerged as an architecture for high-performance computing workloads. Recently, researchers have also focused on the isolation property of a dedicated GPU and suggested GPU-based secure computing environments with several promising applications. Through implementations and security analysis, they have claimed that a dedicated GPU can be leveraged as a secure processor in the presence of a kernel-privileged attacker. However, despite the security analysis conducted by the prior studies, it has been unclear whether a dedicated GPU can support secure execution environment against a kernel-privileged attacker. In this dissertation, we first investigate the security of dedicated GPUs through comprehensive studies on context information for GPU kernel execution. We show that a kernel-privileged attacker can manipulate the GPU contexts to redirect memory accesses or execute arbitrary GPU codes on the running GPU kernel. Furthermore, we introduce the installation method of stealthy rootkit hidden inside GPU kernel execution through GPU context manipulation. Based on our security analysis, we propose a new secure isolated on-chip execution model for the dedicated GPU and a novel defense mechanism supporting the security of the on-chip execution. Our solution operates on Nvidia commodity GPUs without any modification of the hardware or GPU driver and without an additional hardware-abstraction layer. With comprehensive evaluation, we assure that the proposed solutions effectively isolate sensitive information in on-chip storages and defend against known attack vectors from a privileged attacker, supporting that the commodity GPUs can be leveraged as a secure processor.