VSO / Research resources

VSO · Public research resources

Cloud credential and operation evidence research

Synthetic research materials separate signature verification, authentication, operation authority and custody of records.

Public-source review: · Informative, non-normative

Offline evidence checker

A bounded Azure-shaped synthetic record subset separates signature verification from issuer, role, tenant and resource authority. Trust, review scope and evaluation time are caller inputs outside the submitted evidence.

Results are supported, contradicted or unresolved. They are conditional evidence assessments, not execution permissions or an overall safety verdict. Exit zero means a report was produced. See Offline cloud credential evidence checker.

Local collection, restart and portable snapshots

The separate collector retains original bytes of an existing local file and records a signed receipt. It exercises filesystem persistence, restart and retry. Portable snapshots support checks without the original database or private key.

Pin the public key and expected store ID separately. A valid older snapshot does not establish latestness; independently retained checkpoints introduce separate comparison conditions. See Local cloud evidence collector.

Connection, authority and disclosure limits

  • No live Azure connection or collection from an authenticated endpoint.
  • Local possession and a collector signature do not authenticate Microsoft as the source.
  • Successful checks do not establish complete collection, independent custody, power-loss durability or production readiness.
  • The tools neither scan originals for secrets nor authorize their publication.

Choose the experiment before running it

Start with the Python 3.12 offline checker for evidence-content questions. The local collector needs Linux-specific filesystem facilities. Follow each README’s prerequisites and retain separate records for each experiment.

Sources and reviewed revisions

This page summarizes public materials at pinned commits. Creating it is not a new research test run. Historical test records and fixed distributions remain unchanged.