Next-Generation Rowhammer Attacks and Mitigations

• Applicant: Professor Dr. Florian Eduard Walter Adamsky
• Subject Area: Security and Dependability, Operating-, Communication- and Distributed Systems
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 503876675

Wider research context. DRAM stores data in memory cells consisting of capacitors and transistors. These DRAM cells are organized in arrays of rows and columns. Vendors are constantly increasing the density of these memory cells, to optimize for storage capacity, performance, and efficiency. The density is so high that rapidly that reading content of nearby memory rows rapidly can cause bit flips in adjacent memory rows. The exploitation of this so-called Rowhammer effect can undermine memory isolation and, thus, system security. Over the last years, Rowhammer has become a bigger security problem, for two reasons: First, with the density the number of rapid accesses required for an exploit has decreased by a factor of 30. Second, researchers discovered many different ways to exploit Rowhammer for privilege escalation on desktop computers, mobile devices, and even on cloud systems without any software vulnerabilities. Third, prior attempts to mitigate Rowhammer, in particular on commodity systems, have been bypassed by researchers already. Innovation. As outlined above, the Rowhammer effect and its implications are not yet fully understood in many applications and environments. This research project addresses these gaps in our understanding by investigating and developing effective Rowhammer mitigations. Hypotheses. Rowhammer is not an isolated problem but a systemic design flaw and more widespread than known. Rowhammer depends on various real-world environmental influences. DDR5 and GDDR memory is vulnerable to Rowhammer as well. Software-based mitigations can pinpointedly protect targets of privilege-escalation attacks effectively and efficiently. Approach and Methods. We will investigate the scale of the Rowhammer effect to determine whether it is an isolated phenomenon of specific known or yet unknown properties or a systemic design flaw in all modern DRAM. First, we will develop an automatic test suite that helps both researchers and system administrators to test their systems, which we will then use to perform a large-scale study of the Rowhammer effect in the real world. Second, we will investigate Rowhammer in different environments to study the impact on temperature, EM radiation, and aging. Third, we will investigate Rowhammer on graphic cards, since modern graphics cards contain special GDDR memory, which we suspect to be vulnerable as well. Lastly, we will investigate software-based mitigations against Rowhammer to protect page tables, deduplicated pages, and page cache pages. We will develop prototypes to demonstrate the effectiveness of these mitigations.

• DFG Programme: Research Grants
• International Connection: Austria
• Partner Organisation: Fonds zur Förderung der wissenschaftlichen Forschung (FWF)
• Cooperation Partner: Professor Dr. Daniel Gruss