ML Security PapersVulnerabilities in Partial TEE-Shielded LLM Inference with Precomputed Noise
Abhishek Saini, Haolin Jiang, Hang Liu · The State University of New Jersey
Exploits key-reuse in TEE-shielded LLM inference to recover LLaMA-3 8B model weights in 6 minutes and bypass integrity checks
The deployment of large language models (LLMs) on third-party devices requires new ways to protect model intellectual property. While Trusted Execution Environments (TEEs) offer a promising solution, their performance limits can lead to a critical compromise: using a precomputed, static secret basis to accelerate cryptographic operations. We demonstrate that this mainstream design pattern introduces a classic cryptographic flaw, the reuse of secret keying material, into the system's protocol. We prove its vulnerability with two distinct attacks: First, our attack on a model confidentiality system achieves a full confidentiality break by recovering its secret permutations and model weights. Second, our integrity attack completely bypasses the integrity checks of systems like Soter and TSQP. We demonstrate the practicality of our attacks against state-of-the-art LLMs, recovering a layer's secrets from a LLaMA-3 8B model in about 6 minutes and showing the attack scales to compromise 405B-parameter LLMs across a variety of configurations.