Amplifying Latency and Cost of LLM Inference via Rowhammer Attack
LLVMhammer is presented, the first Rowhammer attack that causes LLM inference slowdown by inducing a bit-flip in Triton's compilation pipeline and it is shown that LLVMhammer can also corrupt LLM outputs across different GPU vendors, reducing BERTScore to as low as 25\%, illustrating the broader security implications o...