Good-Enough LLM Obfuscation (GELO)

Anatoly Belikov, Ilya Fedotov

Large Language Models (LLMs) are increasingly served on shared accelerators where an adversary with read access to device memory can observe KV caches and hidden states, threatenin…