Safety researchers from College of Toronto warn on Rowhammer flawOlder GPUs appear to be vulnerableNvidia urges customers to replace as quickly as possibleNvidia is urging customers to use mitigations it supplied in opposition to so-called Rowhammer assaults after new analysis confirmed their potential to trigger critical and stealthy hardware-level compromises.Rowhammer is an exploit of a vulnerability in dynamic RAM (DRAM), the place repeatedly accessing (or “hammering”) a row of reminiscence may cause bit flips in adjoining rows. Consequently, risk actors may bypass safety boundaries, triggering privilege escalations, information tampering, and even denial-of-service states.Though it is a hardware-level difficulty, software-based methods can set off and weaponize the flaw remotely.
You could like
Newer GPUs are safeAlthough recognized for greater than a decade, Rowhammer assaults have first been exploited in 2018, and even then – very not often and in restricted capability – principally on account of their complexity and {hardware} dependencies.Nonetheless safety researchers Chris (Shaopeng) Lin, Joyce Qu, and Gururaj Saileshwar, from the College of Toronto not too long ago printed new analysis demonstrating the sensible use of the flaw:”We ran GPUHammer on an NVIDIA RTX A6000 (48 GB GDDR6) throughout 4 DRAM banks and noticed 8 distinct single-bit flips, and bit-flips throughout all examined banks,” the researchers mentioned. “The minimal activation rely (TRH) to induce a flip was ~12K, per prior DDR4 findings.””Utilizing these flips, we carried out the primary ML accuracy degradation assault utilizing Rowhammer on a GPU.”Signal as much as the TechRadar Professional e-newsletter to get all the highest information, opinion, options and steerage your enterprise must succeed!The “ML accuracy degradation assault” means Rowhammer was used to degrade machine-learning mannequin accuracy, from the standard 80% all the way down to a miserable 1%, utilizing a single bit flip.Nvidia has urged customers to activate the System Stage Error-Correcting Code mitigation, which protects in opposition to Rowhammer on GDDR6 units. The mitigation works by including redundant bits and correcting single-bit errors, sustaining information reliability and accuracy.The listing of affected GPUs is reasonably in depth, and moreover the RTX A6000, consists of a number of Blackwell, Volta, and Turing merchandise.The total listing will be discovered on this hyperlink – however newer GPUs include built-in safety, Nvidia mentioned.Through BleepingComputerYou may also like