Porous GPUS: Side-channel Information Leakage for Secret Recovery
Date: Thu, March 28, 2019
Time: 10:00am
Location: Holmes 389
Speaker: Zhen Hang Jiang
Abstract:
Graphics Processing Units (GPUs) have become the accelerator of choice in the era of big data to support an increasing number of applications. They have been used in security engines where high-volume data is encrypted to ensure confidentiality and signed for integrity. However, the security of GPUs has only begun to receive attention. Despite the challenges posed by massive parallelism and obscure implementation details of GPUs, we have led the first effort in understanding the security implication of the underlying GPU microarchitecture and revealing their side-channel vulnerabilities, often times due to performance features.
In this talk I will present two first-of-its kind side-channel attacks of GPUs, exploiting memory coalescing unit and shared memory bank, respectively. We show, for the first time in the literature, that commercial GPUs are also vulnerable to timing side-channel attacks. We first reverse engineer the vulnerable GPU microarchitectures and identify the timing information leakage. We then take AES (Advanced Encryption Standard) as an example of victims and demonstrate complete key recovery attacks. Our work has exposed security vulnerabilities of GPGPUs and calls on protections against such attacks. There are already several follow-on work of countermeasures to mitigate information from other researchers.
Bio
Zhen Hang Jiang's research interests are in hardware security and side-channel security of hardware-assisted machine learning. His work has been published in several leading computer architecture and embedded systems venues such as IEEE HPCA, IEEE/ACM ICCAD, IEEE ICCD, and ACM GLSVLSI. He is a recipient of the Northeastern University College of Engineering Dean fellowship in 2014 and a recipient of Dissertation Completion Fellowship in 2019. He received his B.S. in Computer Engineering from Northeastern University in 2014 with the highest distinction. He will receive his Ph.D. in Computer Engineering from Northeastern University in Summer 2019.