Build provenance

Build provenance is verifiable metadata about an artifact’s origin. It records what was built, from which source revision, by which build system, using which inputs and parameters — and it is signed, so a consumer can check the claim rather than take it on trust.

It answers a different question from the two artifacts it is usually shipped alongside. An SBOM says what is inside the artifact. A signature says this artifact has not been altered since it was signed. Provenance says this artifact was produced by that pipeline, from that commit — which is what lets you reject a binary that is perfectly well-formed and correctly signed by someone, but did not come from your build system.

The dominant format is in-toto attestations (in-toto is an open framework for securing software supply chains, and the format SLSA specifies for provenance), typically signed with Cosign and distributed alongside the artifact.

For a virtual machine image, provenance is what connects the image you are about to boot in production back to a specific upstream base image and a specific build — so that “where did this AMI come from?” has an answer that does not depend on anybody’s memory.