Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Rowhammer flips bits in DRAM by repeatedly accessing neighbouring rows, letting software corrupt memory it was never granted access to and escalate privilege without any software vulnerability.

What Is a Rowhammer Attack?
Rowhammer is a fault in the physical behaviour of memory chips, not a bug in any program. Repeatedly activating one row of DRAM cells causes electrical disturbance in the rows physically adjacent to it, and with enough activations before the next refresh, bits in those neighbouring rows flip. Since the attacker chooses which rows to hammer and the victim data may sit next door, software can corrupt memory it has no permission to write.
DRAM stores each bit as charge in a capacitor, refreshed on a cycle defined by the standard, typically every 64 milliseconds. As manufacturers shrank cells and packed rows closer together, the electrical coupling between adjacent rows grew until activation of one row could accelerate charge leakage in another. Yoongu Kim and colleagues documented this in a 2014 ISCA paper, testing commodity modules and finding bit flips in a large proportion of those manufactured from 2010 onwards. The finding was presented as a reliability problem.
In March 2015, Google Project Zero showed it was a security problem. Mark Seaborn and Thomas Dullien demonstrated two exploits: escaping the Native Client sandbox, and obtaining kernel privileges on Linux by flipping a bit in a page table entry so that a writable page gained access to a page table itself, which converts one bit flip into control of all physical memory.
Three steps recur across almost every Rowhammer exploit.
Variants extended the reach considerably. Drammer, in October 2016, achieved root on Android handsets from an app with no permissions. GLitch in 2018 triggered flips from JavaScript using the GPU. ECCploit showed in 2018 that error-correcting memory slowed the attack rather than preventing it. TRRespass in 2020 defeated Target Row Refresh, the in-DRAM mitigation vendors had introduced, by hammering many rows at once. Blacksmith followed in 2021 with non-uniform hammering patterns, and ZenHammer in March 2024 demonstrated flips on DDR4 and DDR5 systems using AMD Zen processors.
The attack requires code execution on the target, so it is a privilege escalation and isolation-breaking technique rather than an entry point. Its significance is where isolation is the whole product: a cloud host running untrusted tenants, a browser sandbox, or a trusted execution environment relying on memory it does not physically control. Cross-tenant attacks in public cloud have been demonstrated in research conditions and are hard to execute against modern hypervisors with memory deduplication disabled, but the threat model is real enough that providers invest in it.
For a typical enterprise estate, Rowhammer is not the risk that should keep anyone awake. It needs local code execution, extensive profiling time, and specific memory hardware, and simpler privilege escalation routes are usually available. It belongs in the risk register for multi-tenant infrastructure, high-assurance systems, and long-lived embedded equipment.
There is no patch, because the defect is physical. ECC memory detects single-bit errors and corrects them, which raises the effort required and, more usefully, produces machine check events that can be monitored as a detection signal. Increasing the refresh rate reduces the window for disturbance at the cost of performance and power. DDR5 adds on-die error correction and improved refresh management, and newer standards tighten the activation limits that make hammering effective, although research has continued to find patterns that work.
Practically: buy server hardware with ECC and monitor correctable error counters, keep hypervisor memory deduplication disabled in multi-tenant environments, apply firmware and microcode updates that adjust refresh behaviour, and for the highest-assurance workloads prefer dedicated hosts over shared tenancy. Our Hardware Penetration Test Services cover memory-level attacks for organisations building devices where physical isolation is part of the security claim.
Only in limited research scenarios. Most exploits need code execution on the target, delivered through an application, a sandboxed browser context, or a virtual machine on shared hardware. Work such as Throwhammer and Nethammer showed flips triggered through high-rate network traffic over RDMA, but these require unusual configurations. Treat it as a privilege escalation technique rather than an initial access route.
It raises the difficulty without eliminating the attack. Standard ECC corrects a single-bit error per word and detects two, so an attacker needs multiple precise flips within one word, which the ECCploit research demonstrated is achievable with enough time. The more valuable property is detection, because correctable error counters rise sharply during hammering and can be monitored.
No, though it is harder to attack. DDR5 adds on-die error correction and refresh management improvements, and later standards tighten activation limits. Research including ZenHammer in 2024 has still produced bit flips on DDR5 platforms. The trend is that each mitigation raises the cost and each generation of research finds new access patterns that work within it.
For most workloads, no. For systems where tenant isolation is itself the security control, such as processing of special category data or key management, dedicated instances or bare metal remove the shared-DRAM concern entirely. Major providers already disable memory deduplication and apply hardware-level mitigations, and their security documentation is the right place to confirm the current position.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.