Identity is never self-certifying
Why is authentication the problem of proving a secret without creating a secret?
Identity and access protects a decision, not data directly. Everything reduces to one question: may this principal perform this action on this object, right now? — and then trusting the answer. The mechanisms divide into three sub-questions:
- How do you know who is asking? (authentication)
- How do you compute the answer? (authorization models)
- Where do you put the machine that enforces it? (boundaries, grants, vending, delegation — Enforcement points)
The governing constraint is that identity is never self-certifying. Bits carry no proof of where they came from. A request that says "I am the risk-reporting job" looks exactly like a false request that says the same, unless it comes with evidence that only the real principal could produce, checked by a party you decided in advance to believe. Every authentication system ends in a root of trust that was set up outside the system (by a person, on a day, with a ceremony). No amount of cryptography moves that end point. It only makes the chain to it shorter.