Evidence-Bounded Change Assurance for Cloud Security
Know what a security change will actually break — and know when the evidence isn’t enough.
Before you revoke a permission, Cyber-Circuit evaluates the proposed change against a real observed operation and a captured reference authorization state.
Current AWS workflow: ESTABLISHED_AFFECTED when the evidence establishes impact. UNKNOWN when the evidence cannot establish the effect.
A permission can look unused and still be carrying production.
A permission’s presence tells you it exists — not whether a real production operation depends on it.
Revoke uncertain access and production can break. Keep every grant in place and unnecessary exposure accumulates.
“For this observed operation, using this captured authorization state as the evaluation reference, what does the evidence establish about this proposed withdrawal?”
Cyber-Circuit answers this question without converting missing evidence into certainty.
Method
Evidence before assertion.
01
Observe
Anchor the evaluation to a real CloudTrail-observed operation.
02
Capture
Capture the acting principal’s relevant authorization state as the reference.
03
Define
Identify the exact permission withdrawal being considered.
04
Simulate
Run the counterfactual through live AWS IAM policy simulation.
05
Verify
Immediately re-collect the acting principal and compare against the reference state.
06
Bound
Return the strongest conclusion the evidence supports — or UNKNOWN.
State verification is a bounded post-evaluation check, not a transactional guarantee that IAM could not change during the simulator call.
Current validated scope
- AWS IAM
- CloudTrail-observed S3 GetObject
- IAM role authorization
- Exact-object inline role-policy grants
- Evidence-backed permission-withdrawal evaluation
Evidence
The conclusion comes with its working.
The CloudTrail record anchoring the operation.
The captured state used as the evaluation reference.
An identifier for the captured authorization state used as the evaluation reference.
The authorization relationships relevant to the proposed withdrawal.
Evidence captured from live IAM policy simulation.
Immediate post-evaluation re-collection compared against the reference.
The missing or inconsistent context behind an UNKNOWN result.
Validation
Small cases. Inspectable evidence.
In the controlled Phase 5B experiment, the observed operation depended on one necessary exact-object grant. With that grant removed, the operation was subsequently denied. IAM stability during the original experiment was maintained operationally.
Independent authorization paths kept the evidence from supporting a stronger conclusion.
Required conditional context was absent, so the evidence could not establish the effect.
Evaluate the change before you make it.
Bring Cyber-Circuit one real observed operation and one proposed permission withdrawal. Get an evidence-bounded evaluation showing what can be established, what cannot, and why.
No universal safety claims. No opaque confidence scores. No inferred certainty across missing evidence.