Trust by Proof, Not by Popularity - Verifying Container Provenance

You picked that base image because it had millions of pulls and an “official” badge. That’s the whole security review—popularity stood in for proof. But download counts tell you nothing about what’s inside the image, how it was built, or who built it, and recent supply-chain attacks (SolarWinds, Codecov, ua-parser-js) all rode through widely trusted artifacts. 

Here’s why this resists an easy fix: popularity is a social signal, and you can’t scan your way back to a build process that was never recorded. Once you pull a pre-built binary, the provenance is already gone—there’s no after-the-fact tool that reconstructs how it was made or proves it wasn’t tampered with. 

I’ll show how to replace reputation with cryptographic proof: source-based builds with SLSA provenance, signed attestations, and verification enforced at the cluster edge with a policy controller and admission webhook—so an image without valid provenance simply can’t deploy. I’ll walk through the failure mode of trusting an “official” image and the moment a verification gate would have caught it. 

You’ll leave with an admission-control verification pattern and a short checklist for vetting any third-party image before it enters your cluster—both buildable on free, open-source tooling this sprint. 

Differentiation: provenance talks usually stop at “sign your images.” This one is about the consumer side—how to refuse images you can’t verify, enforced automatically at admission, not by policy you hope people follow. 

1. How to sign and verify container images using Sigstore/Cosign — step-by-step CLI commands to start using immediately, no commercial tooling required.

2. What SLSA Level 4 provenance provides and how to enforce it via a Kyverno ClusterPolicy in Kubernetes — copy-paste policy included.

3. A trust-model decision framework: concrete criteria for when you are trusting by cryptographic proof versus trusting by popularity — and a checklist to audit your current image intake process.