[Seminar] Understanding and Enhancing Microarchitecture Security in the Era of AI and Emerging Hardware
Friday, February 10, 2023
11:00 am - 12:00 pm
Speaker
Fan Yao, Ph.D.
University of Central Florida
Location
PGH
232
Abstract
Recent developments of adversarial exploitation rooting in hardware (e.g., microarchitecturalattacks and rowhammer) have forcibly opened a new chapter for computing system security, highlighting the fact that the underlying hardware (i.e., internal threats) is at the center of future attacks and defenses. The security prospect of future computing is even more concerning with the burgeoning of artificial intelligence and machine learning (ML) techniques, which are ubiquitously integrated into our daily lives to perform security-sensitive tasks (e.g., autonomous driving, recognition and access control).
In this talk, I will first present our investigations on novel attacks that leverage hardware vulnerabilities(i.e., DRAM memory fault attacks and side channels) to compromise critical security primitives of state-of-the-art ML models (i.e., integrity and confidentiality). These attacks for the first time show the practicality of model tampering and model extraction by directly targeting model parameters (i.e., weights), which opens a new research dimension in adversarial machine learning. Our work represents the first set of research endeavors to understand the security of ML systems from the lens of real-world hardware. In addition, I will also present our recent work that proposes secure architectures with efficient processor-side support to offer holistic data confidentiality and integrity protection for systems with emerging memory and storage devices.
About the Speaker
Dr. Fan Yao is an assistant professor in the department of Electrical and Computer Engineering at the University of Central Florida. He received his Ph.D. degree in Computer Engineering from the George Washington University (GWU), and the B.E. degree from Huazhong University of Science and Technology. Fan is currently directing the Computer Architecture and Systems Research Lab (CASRL) at UCF. His research interests broadly lie in the areas of computer architecture, systems and security. Hisresearch focuses on architecture/hardware security (e.g., microarchitectural attacks and defenses), AI/ML hardware security (e.g., fault attacks and side channels), efficient and secure integration of emerging hardware technologies (e.g., non-volatile memory), and energy-efficient computing. 贵补苍鈥檚 research projects are actively supported through several competitive federal grants. His research group constantlypublishes high-quality papers in top conference and journal venues (e.g., MICRO, HPCA, USENIX Security, S&P, EuroS&P, HOST, DSN, IEEE TC, IEEE Micro and IEEE TIFS). Fan has been serving in the technical program committees and organizing committees of several major international conferences in the research field, including HPCA, MICRO, ISCA and DAC. 贵补苍鈥檚 work has been selected as Top Picks in Hardware and Embedded Security in 2019. He has received the GWU Best Ph.D. Dissertation Award in 2018.
