Passive network observer
TLS, signed packets, message encryption, WebRTC DTLS/SRTP, size buckets, and generic Web Push protect content. Timing, endpoints, traffic volume, and approximate size remain observable.
TLS, signed packets, message encryption, WebRTC DTLS/SRTP, size buckets, and generic Web Push protect content. Timing, endpoints, traffic volume, and approximate size remain observable.
The node can drop, delay, replay, reorder, selectively route, or measure traffic. Device bindings, packet signatures, ratchet replay protection, recipient checks, delete capabilities, and local downgrade pins protect integrity/confidentiality. Availability and traffic analysis remain residual risks.
Every device key object must be signed by the paired identity and pinned locally. Existing device IDs are immutable. Unknown, changed, or unbound keys fail closed.
The ratchet handles out-of-order messages and rejects ciphertext replay/tampering. Node key claims are atomic, consumed keys are tombstoned, and request replay returns the original claim response.
A member can preserve plaintext they legitimately receive, take screenshots, re-share content, flood a group, or compromise their own endpoint. Cryptography cannot revoke already received plaintext.
A tenant administrator can add, remove, disable, or reassign users and cause denial of service through policy changes. The administrator cannot decrypt content through supported control APIs unless it assigns an account it controls as a member and that account is subsequently provisioned.
This is the highest confidentiality risk. CSP, same-origin assets, escaped dynamic text, no remote cryptography code, dependency pinning, and local kill reduce exposure. They do not protect against a fully compromised origin, malicious browser extension, unlocked OS account, memory inspection, or keylogger.
Dependencies are exact-pinned with a lockfile and notices. WASM is served locally. A compromised signed dependency or build host remains a residual risk.
Subscription writes require a signed account/device and mailbox binding. Endpoints are HTTPS and browser-provider allowlisted. Payload content is constant and encrypted. Provider compromise still exposes endpoint and timing metadata.
Packet size limits, bounded batches, IP and signed-sender rate limits, indexed mailbox lookup, storage quotas, expiry, and bounded community relay storage provide controls. An adversary can still generate many identities or attack infrastructure upstream.
Tenant and operator APIs are separate from ciphertext keys. Admins can disable users, alter policy, and cause denial of service. They cannot decrypt content through supported APIs. A stolen administrator token can deny access or replace organization membership, but cannot decrypt retained endpoint history through the recovery flow.