KEM-PRTL-001 · Protocol Charter
The Kemet Secure Messaging Protocol — end-to-end confidentiality, authentication, and session integrity for private communications on the Network.
We establish KSMP as the authoritative encrypted messaging instrument of the Network. The Organization operates an encrypted relay and durable queue — not a custodian of plaintext. Server infrastructure may observe routing identifiers, timestamps, envelope size classes, and protocol version markers. It cannot forge valid authenticated ciphertext or recover message bodies without client key material.
Identity — long-lived cryptographic node bound to a Vault-derived key domain Device — one installation, one device key domain, auditable enrollment Conversation — deterministic thread namespace across participant devices
Sessions progress through defined states: NEW → ACTIVE → DESYNCED → REKEY_REQUIRED → ACTIVE. We require explicit re-establishment after repeated decrypt failures. Clients that permit insecure fallback when prekey inventory is exhausted are non-compliant with this charter.
Message bodies and attachment metadata are encrypted on the sender device. Routing envelopes carry immutable header fields authenticated under AEAD associated data. Attachments are encrypted client-side before upload; content keys travel inside message plaintext, never in cleartext at the relay layer.
Fan-out is per recipient device, not once per user. Delivery and read receipts are idempotent, monotonic events keyed by message identity. The Organization does not use receipt metadata to reconstruct message content.
KSMP v3 extends the messaging contract with server-authoritative ordering, tickle-only push, and a unified inbox replay model. Both generations uphold the same plaintext boundary: the relay never holds decryptable content.
For implementer-level session requirements, wire events, and checklists, see the KSMP developer reference.
Disclosure Line
This charter publishes the contractual behavior, guarantees, and governance boundaries of the protocol. It does not publish anti-abuse scoring internals, exploit-sensitive thresholds, or operational topology details that materially improve adversarial optimization against the Network.