Skip to content

[feature] Unifying SBOM Representation and Trust Across Multiple Equivalent Image References #2387

@robert-cronin

Description

@robert-cronin

In container ecosystems, a single image (identified by a digest) can appear in multiple registries under different tags or repositories. For instance, an image at registryA.com/base:1.0 can be mirrored at registryB.com/cache/base:1.0. Although these references differ, the underlying image content is identical. A single, signed SBOM should logically apply to all such references without needing to be regenerated (which would break its original signature and trust).

Currently, GUAC typically associates one SBOM per subject (e.g., one image reference), making it hard to unify multiple references of the same image or multiple SBOM formats (e.g., SPDX and CycloneDX) for the same underlying content. Additionally, SBOMs aren’t first-class entities with their own trust or provenance data, limiting the ability to differentiate a trusted SBOM from a suspicious or forged one.

Use Case Example:

1. Multiple Registry Locations, One Image:

  • Consider an official base image at registryA.com/base:1.0 described by a trusted SBOM.
  • Another registry provides a pull-through cache of this same digest at registryB.com/cache/base:1.0.
  • In GUAC today, there’s no simple way to say: These two references represent the same underlying image, so we should link them to the same trusted SBOM.

2. Different SBOM Formats, Same Content:

  • If the same image from above also has a CycloneDX SBOM and an SPDX SBOM, both describing identical content, GUAC cannot directly represent these two documents as equivalent descriptions.
  • Users can’t query: “Show me all SBOM representations of this image” and see them unified.

3. Trust and Attestation at the SBOM Level:

  • Suppose a malicious actor introduces a forged SBOM claiming to describe the same package.
  • Without modeling SBOMs as distinct nodes, GUAC can’t easily attach attestations, cryptographic proofs, or trust metadata to that SBOM and differentiate it from the known-trusted one.

This issue aims to explore ways to unify image references, preserve trust in original SBOMs, and enhance the expressiveness of SBOM modeling in GUAC.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions