Releases#
Release v4.0.5#
Date: April 2, 2025
Introduced
/v4
APIs for GPU and Switch Attestation/v4
APIs support both 2.0 and 3.0 claims. The 3.0 claims offer greater granularity by reducing bundled claims and focusing on individual ones. For details on the claims, see the NvTrust Troubleshooting Guide.
Fixed security vulnerabilities
Release v3.5.1#
Date: February 6, 2025
Fixes for regular cache invalidation
Release v3.5#
Date: January 29, 2025
Integration with UAM for authorization workflow. Note: The request would be processed by the service irrespective of the outcome of authorization result
Fixed security vulnerabilities
Support for Blackwell architecture
Distroless container version upgrade
Release v3.4#
Date: January 13, 2025
Health endpoint has been added. A new comprehensive health endpoint has been introduced to provide detailed insights into the status of dependencies and connected services.
Fixed security vulnerabilities
Performance improvements for multi-device attestation
Release v3.3#
Date: December 5, 2024
Change attestation behavior for pre-release drivers (aka “PID drops”). Attestation would previously pass if a RIM certificate had a certificate hold revocation status. This revocation status is used to indicate that a RIM corresponds to a PID drop. To opt-in to PID drops, clients must pass the header
X-NVIDIA-OCSP-ALLOW-CERT-HOLD
with the valuetrue
. This change applies to all attestation endpoints.Performance improvements for multi-device attestation.
Release v3.0#
Date: October 22, 2024
Stabilized multi-GPU and multi-switch attestation endpoints under
/v3
APIs. Changes from the/v2
attestation APIs are:Improve consistency of claim names and advance claims version to 2.0. For details on the 2.0 claims, see the NvTrust Troubleshooting Guide.
The
/v3/attest/switch
now requires anarch
field to indicate the switch architecture.Use more lightweight encoding scheme for device evidence. Evidence fields are now expected to contain a base-64 encoded SPDM request and response.
Update
evidences
field name toevidence_list
.Ensure device subtokens are consistently indexed in all scenarios. In certain failure modes, error tokens were returned under the wrong key (
gpu{index}
instead of the expectedgpu-{index}
).
Move OpenAPI specification from
/v3/api-docs
to/api-docs
and improve API documentation layout.
Release v2.0#
Date: August 27, 2024
Migrated to a more robust and fault-tolerant platform, improving performance and reducing error rates.
Deprecated the
/v1/attestation-docs
API.
Release v1.1.0#
Date: August 7, 2024
Introduced the
/v2
API for multi-GPU attestation. The/v2
API also supports multi-nvSwitch attestation, used for attesting nvSwitches. Note: This is an EA (Early Access) release and does not include the release of corresponding nvTrust / SDK changes.Remote GPU / nvSwitch attestation responses for multi-GPU / multi-Switch configurations will follow section “A.2.3. JSON-encoded Detached EAT Bundle” of the RATS specification. The response will include:
An overall attestation JWT claim that reflects the overall attestation result.
Individual GPU tokens represented as detached EAT bundles:
The order of GPU / nvSwitch claim bundles will match the order of the evidence list.
A unique key will be created for each result, representing the individual GPU/Switch claims (e.g.,
gpu-0
orswitch-0
).
Release v1.0#
Date: July 29, 2023
General Availability (GA) release of the NVIDIA Remote Attestation Service (NRAS).
Provides GPU attestation support for H100 GPUs.
Generates signed EAT (Entity Attestation Token) as the attestation response.