Tamper-evident metadata

Tamper-evident metadata is signed metadata that allows a consumer to detect whether it has been modified. The signer (the system that wrote the metadata) computes a cryptographic hash or signature over the canonical form of the metadata and stores the signature alongside it in the file. A consumer reads the metadata and signature, recomputes the hash or signature using the same canonicalization rules, and compares. If they match, the metadata hasn't been tampered with. If they differ, tampering is evident.

How it works

LRFS uses HMAC-SHA256: the writer hashes the canonical XMP content with a shared secret key, producing a signature. The signature is stored in the XMP packet. A consumer with the same key recomputes the hash; match means integrity, mismatch means tampering. For public verification, ed25519 is used: the writer signs with a private key, the consumer verifies with a public key, and anyone with the public key can cryptographically prove the signature is valid.

Why 'evident' not 'proof'

Tamper-evident means you can detect modification, not prevent it. An attacker can rewrite the entire file and compute a new signature — they just can't modify the metadata silently. Consumers who verify signatures gain confidence in integrity; consumers who skip verification are at risk. Tamper-evident is about transparency and auditability, not absolute protection.