LATIDIA · Ciberseguridad

Amplificación de la latencia y el coste de la inferencia LLM a través del ataque Rowhammer

arXiv: 2610.30798v1Tipo de anuncio: nuevo Resumen: Los modelos de lenguaje grandes (LLM) logran una alta eficiencia de inferencia utilizando núcleos de GPU optimizados generados por marcos de compilación justo a tiempo (JIT) como Triton. Triton com

La noticia

arXiv:2610.30798v1 Announce Type: new Abstract: Large language models (LLMs) achieve high inference efficiency using optimized GPU kernels generated by just-in-time (JIT) compilation frameworks such as Triton. Triton compiles GPU kernels through an MLIR/LLVM-based pipeline. We investigate whether a Rowhammer-induced bit-flip in the CPU's DRAM can degrade GPU-based LLM inference performance while preserving output correctness. We present LLVMhammer, the first Rowhammer attack that causes LLM inference slowdown by inducing a bit-flip in Triton's compilation pipeline. A bit-flip in a performance-critical optimization pass causes Triton to generate a functionally identical but suboptimal GPU kernel. Since Triton caches the generated kernel on disk, a single successful bit-flip causes persistent slowdown across subsequent inference