GPUTHOR ATTACK BREACHES NVIDIA GPU SECURITY

Security researchers have disclosed GPUThor, a hardware vulnerability that exploits NVIDIA's GPU architecture. The attack targets ECC memory protection mechanisms designed to detect and correct bit errors in GPU memory. By inducing controlled bit flips through repeated memory access patterns, GPUThor can corrupt data structures and gain elevated system privileges. The vulnerability affects NVIDIA's GPU lineup and potentially impacts data center environments relying on GPU acceleration. The attack requires local access to the target system but poses significant risks for cloud GPU services and shared GPU environments. NVIDIA has been notified of the vulnerability. Organizations using NVIDIA GPUs should monitor for security updates and consider access controls limiting local user privileges. The discovery underscores ongoing challenges in hardware security, particularly with Rowhammer-class vulnerabilities that exploit the physical properties of memory cells. Researchers continue identifying new variants affecting various processor architectures.