Epistemology of Rowhammer Attacks: Threats to Rowhammer Research Validity

The Rowhammer effect is a disturbance error in DRAM that attackers can trigger from software. The first publication on Rowhammer in 2014 evaluated 129 Dual In-line Memory Modules (DIMMs) on an FPGA and showed that 110 DIMMs are affected, indicating that Rowhammer is a widespread issue. However, until now, no case outside of academia is known in which Rowhammer was used for attacks, indicating a stark discrepancy between the attention Rowhammer receives and its real-world relevance.

This paper systematically analyzes 32 offensive Rowhammer papers, including 48 experiments. However, we avoid finger-pointing but identify six threats to the validity and relevance of Rowhammer research results and give multiple examples. The threats include small sample sizes, overestimated attacker capabilities, unrealistic attack scenarios, non-comparability of the results, age and wear of hardware, and sub-optimal attack performance metrics. Additionally, we provide recommendations with detailed justification to the scientific community to mitigate those threats: (1) pre-experimental testing of DIMM integrity, (2) increasing and broadening the DIMM sample size, (3) expanding reproduction studies of published work, (4) defining attacks in real-world conditions and distinguishing them from theoretical ones, (5) publishing DIMM manufacturing data, (6) documenting DIMM wear and, (7) leveraging multiple metrics for bit flip evaluations.