Encryption makes data as secret as the key

Why does protection shrink the secret rather than remove it?

Identity and access and Enforcement points assumed that you own the chokepoint (the single place all reads must pass). Protection is what you build when you do not own it, because the adversary already holds the bytes. Examples: the stolen drive, the copied backup, the cloud operator, the compromised host, the analyst who exported data to a laptop. The question changes from who may read this? to what can be read by someone who has already read it?

The governing constraint is a resource asymmetry: you cannot guard a petabyte, but you can guard 32 bytes. Every mechanism in this article is a scheme that collapses the confidentiality of a large object onto a small secret. That is a real gain, but it has a condition that is easy to state and hard to live with: the small secret must be available everywhere the data is legitimately used, at the rate the data is used. A key that is truly unavailable protects nothing, because it also serves nothing.

So the whole article is a negotiation over one number: how much structure of the original data survives the transformation.

• Zero structure survives: perfectly secure and perfectly useless.
• All structure survives: queryable and unprotected.
• Everything real sits in between, and the position you choose is exactly what you have leaked.