Lion Cipher Lab

Build an AWS data perimeter with SCPs, RCPs and VPC endpoint policies (offline evaluator, measured)

Why it matters

Most cloud data theft does not need an exotic exploit. It needs one of three things: a valid identity copying your data to a place you do not own, an outside identity reaching in through a policy that grants too much, or stolen credentials used from somewhere you would never expect. AWS's answer is a data perimeter -- a small set of organization-wide Deny rules that hold no matter how careless any single team's policy is. The hard part is not writing the Deny. The hard part is knowing who each guardrail actually applies to, which condition keys go missing on which network paths, and which…

What you'll learn

Steps in this lab

  1. What a data perimeter is, in three questions
  2. The five policy types, and exactly who each one applies to
  3. Condition keys that are sometimes missing, and why that is the whole game
  4. Get set up (from zero)
  5. Build the engine: who applies, what exists, then deny-allow-grant
  6. Build the toy organization, 21 requests and 9 layers
  7. Run it: the measured before/after table
  8. Layers 1-2: trusted resources (SCP with aws:ResourceOrgID)
  9. Layers 3-5: trusted identities (RCP with aws:PrincipalOrgID), and the service confused deputy
  10. Layers 6-7: expected networks (SCP with aws:SourceIp and aws:SourceVpc)
  11. Layer 8: the VPC endpoint policy catches the request nobody else sees
  12. What survives all eight layers, and what a perimeter never sees
  13. Where this model is simpler than AWS
  14. How a professional rolls this out for real

Open the full interactive lab →