#dram vulnerability

Rowhammer Backdoor Injection Attack During Inference: A New⭐8

A new attack from Northeastern University exploits Rowhammer to inject backdoors into neural networks during inference, threatening AI integrity.