
"The attacks, known as GDDRHammer and GeForce/GeForge, exploit vulnerabilities in GDDR6 GPU memory to gain arbitrary read and write access, ultimately allowing attackers to take control of the host CPU and system memory."
"Unlike earlier GPU-focused attacks that primarily impacted application behavior, these new techniques demonstrate end-to-end compromise capabilities, bridging the gap between GPU and CPU memory spaces."
"Research shows that attacks like GDDRHammer can generate large numbers of targeted bit flips, over 100 per memory bank in some cases, while bypassing existing GPU protections."
Researchers have identified new Rowhammer attacks, GDDRHammer and GeForce/GeForge, that exploit vulnerabilities in GDDR6 GPU memory. These attacks allow arbitrary read and write access, leading to potential full system compromise. Unlike previous GPU attacks that affected application behavior, these techniques enable end-to-end system control. By inducing bit flips in GPU memory, attackers can manipulate memory mappings, gaining unauthorized access to system memory. The risks are particularly pronounced in shared environments like cloud infrastructure and AI training platforms, where GPUs are commonly utilized across multiple workloads.
Read at InfoQ
Unable to calculate read time
Collection
[
|
...
]