Skip to content

Latest commit

 

History

History
41 lines (26 loc) · 2.09 KB

README.md

File metadata and controls

41 lines (26 loc) · 2.09 KB
Raw

Intel® Trust Authority Python NVIDIA* H100* GPU Adapter

· 09/17/2024 ·

The Intel Trust Authority Client for NVIDIA* H100 GPU is a Python package for collecting evidence for attestation from a NVIDIA H100 GPU. This library uses the NVIDIA Attestation SDK (https://github.com/NVIDIA/nvtrust/tree/main/guest_tools/attestation_sdk) for H100 GPU Evidence generation. This GPU adapter is used with the Intel Trust Authority connector for Python to request an attestation token and verify the same.

This version of the GPU adapter works with on-premises Intel® Trust Domain Extensions (Intel® TDX) and NVIDIA H100 Confidential Computing enabled platforms. A future version may support cloud-based platforms.

The GPU adapter can be used to attest only a NVIDIA H100 GPU, but the primary use case is a combined attestation of both the Intel TDX trust domain and the NVIDIA H100 GPU. The GPU adapter collects evidence from the GPU, and the Intel TDX adapter collects evidence from the trust domain. The connector combines the evidence from both adapters and sends it to Intel Trust Authority for verification. If successful, the response is an attestation token (JWT) that can be used to verify the integrity of the platform.

For more information, see GPU Remote Attestation in the Intel Trust Authority documentation.

Requirements

  • Use Python 3.8 or newer.
  • Ubuntu 24.04 with kernel version 6.8 or newer.
  • NVIDIA H100 GPU

Usage

To create a new NVIDIA GPU adapter and use the adapter to collect evidence from Intel TDX and NVIDIA H100 Condidential Computing enabled platform.

#Create a new GPU adapter
adapter = GPUAdapter()

#Use this adapter to get evidence
evidence = adapter.collect_evidence(nonce)
if evidence == None:
    return None #error condition

License

This source is distributed under the BSD-style license found in the LICENSE file.



* Other names and brands may be claimed as the property of others.