RHAT: Efficient RowHammer-Aware Test for Modern DRAM Modules
Mohammad Farmani, Mark Mohammad Tehranipoor, Fahim Rahman · 2021
In recent times, the family of rowhammer attacks have posed major security threats. Assessing the rowhammer vulnerability of DRAM modules during post-manufacturing test as well as system development and in-field deployment is a crucial step towards detecting and mitigating it. Detection of rowhammer vulnerable cells is quite challenging due to random physical factors and memory constructions, and very exhaustive leading to time and cost overhead. In this paper, we propose an efficient test framework, called RHAT, to address this concern. RHAT can be employed for both manufacturing test and in-field (deployment) test. It first develops a non-invasive approach to efficiently reverse engineer the address scrambling on DRAM. It then identifies critical vulnerability features by spatial correlation analysis of memory cells. Finally, it offers a fast test algorithm to detect rowhammer-vulnerable cells using smaller representative memory array based on the correlation analysis. We implemented and validated RHAT on DIMM samples from three different DRAM vendors showing significant correlation with satisfactory test coverage and time.