Intel TDX enables bilateral protection for private LoRA inference
Original title: Bilateral Protection for Private LoRA Inference with Intel TDX
Intel TDX system protects client inputs, outputs, and proprietary LoRA parameters during cloud inference via remote attestation and verified byte consumption.
Cloud services using private LoRA must safeguard inputs and outputs alongside owner parameters. The system connects approved application identities to artifact consumption inside a trust domain. Clients and owners verify binding to TLS keys before releasing decryption keys, with all inference executed in Intel TDX.
- Experiments confirm identical token sequences in 400 dialogues
- Steady-state service-to-native time ratio is 1.0094 on TDX hardware
- Fixed read-only image rejects controlled modifications before secret release
From the original article
1 min read at arxiv.orgCloud services using private low-rank adaptation (LoRA) must protect client inputs and outputs alongside owners' proprietary parameters and execute the approved adaptation. We design and implement a system for bilateral protection of private LoRA inference using Intel Trust Domain Extensions (TDX), connecting approved application identity to actual artifact consumption and adapter activation.
By Xiaodong Li · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.