NVIDIA/nvtrust

What it solves

It provides the necessary software tools to implement and validate "Trusted Computing" on NVIDIA hardware. Specifically, it addresses the need for attestation—verifying that the hardware and software environment is secure and has not been tampered with—which is essential for confidential computing environments.

How it works

The project offers a suite of tools divided by their role in the system:

• Guest Tools: Includes an Attestation SDK (Python) for integrating GPU and NVSwitch attestation into applications, and a PPCIE Verifier for multi-GPU setups using Protected PCIe to maintain security while allowing plain-text NVLink traffic.
• Host Tools: Provides utilities to configure GPU Confidential Computing modes and scripts to launch Confidential VMs.

Who it’s for

Developers and system administrators building secure, confidential computing environments on NVIDIA GPUs who need to verify the integrity of their hardware and software stack.

Highlights

• Attestation SDK: Simplifies the implementation of local and remote attestation workflows.
• Confidential VM Support: Includes host-side tools and KVM scripts for launching secure virtual machines.
• ** uma Protected PCIe (PPCIE) Verification**: Supports multi-GPU confidential computing setups.
• Comprehensive Documentation: Provides guides for moving from proof-of-concept to production deployment.