Hey Golang community,
I’m excited to share that Confident Security is open-sourcing go-nvtrust, a Go library for handling confidential computing attestation on NVIDIA GPUs and NVSwitches. Inspired by NVIDIA’s nvtrust, it provides a straightforward way to collect attestation evidence, perform remote verification via NRAS, and includes Go bindings for libnvidia-nscq.
This library supports Hopper (H100, H200) and Blackwell GPUs, as well as LS10 NVSwitches.
Why does this exist?
We needed a Go-native library rather than NVIDIA’s nvtrust. This library allows us to verify the NVIDIA GPU in use is authentic, untampered, and in confidential mode.
Key Features
- Go bindings for libnvidia-nscq (NVSwitch attestation).
- Integration with NVML via go-nvml for GPU attestation.
- Simple API for evidence collection and NRAS verification.
- Designed for tamper-proof workflows in confidential computing.
Get Started
Repository: GitHub - confidentsecurity/go-nvtrust: go version of https://github.com/nvidia/nvtrust (Apache-2.0 license)
Check out the README for quick start guides, API reference, and examples. We’d love feedback or contributions - has anyone here worked on GPU attestation in Go? How are you securing your NVIDIA setups?
Cheers,
Confident Security