Abstract

Modern DRAM is increasingly susceptible to disturbance effects in which carefully crafted memory-access patterns can induce errors in nearby cells. RowHammer is the best-known example, but rather than behaving simply as random faults, disturbance-induced bit flips can exhibit meaningful spatial structure tied to the underlying memory organization. In this talk, I will present our research on understanding and exploiting this structure from several perspectives. I will discuss spatial characterization of RowHammer behavior across commercial DRAM devices, methods for reducing the cost of vulnerability testing, and convolutional neural network models that predict vulnerable memory regions from partial observations. I will then show how the same disturbance characteristics can be repurposed for hardware security through DRAM-based device identification and physically unclonable functions. I will conclude with ongoing directions involving DDR5, RowPress and other memory-disturbance phenomena, and opportunities for Florida Tech students interested in hardware security, machine learning, computer architecture, and secure systems.

About the Speaker

Mohammad Farmani is an Assistant Professor in the Department of Electrical Engineering and Computer Science at the Florida Institute of Technology. He received his Ph.D. in Electrical and Computer Engineering from the University of Florida, where his dissertation focused on RowHammer vulnerability analysis in DRAM. His research focuses on hardware security, DRAM disturbance vulnerabilities, secure memory architectures, physically unclonable functions, and hardware-assisted cybersecurity. His recent memory-security research has been presented at IEEE HOST, IEEE VTS, IEEE ISVLSI, and DRAMSec. Prior to joining Florida Tech, he was an Assistant Professor of Cybersecurity Engineering at Florida Polytechnic University.