Threat Model

Assets

The primary assets considered by this specification are:

  • confidentiality of data and control traffic when secure transports are used,

  • integrity of transport/session/network/data messages,

  • peer identity during session establishment,

  • correctness of local authorization decisions,

  • availability of established sessions and discovery flows.

Trust Boundaries

Zenoh deployments typically cross the following trust boundaries:

  • the local node process and its local configuration,

  • authenticated peers within an administrative domain,

  • unauthenticated or partially trusted peers reachable on the transport network,

  • intermediate networks and links that may be observed or modified unless protected by a secure transport.

In-Scope Threats

This specification, together with secure transport and local policy enforcement, is intended to help address:

  • eavesdropping on links protected by TLS/QUIC or equivalent secure transports,

  • active tampering with traffic on authenticated secure links,

  • impersonation during session establishment when transport- or session-layer authentication is enabled,

  • unauthorized publish/query/declaration activity rejected by local access-control policy,

  • replay of stale session-establishment state through the INIT/OPEN cookie mechanism.

Out-of-Scope Threats

The following are out of scope for the base wire specification:

  • physical compromise of devices or the trusted local process,

  • operating-system, runtime, or side-channel attacks,

  • IP-layer or link-layer denial of service,

  • traffic analysis on otherwise encrypted links,

  • key management, PKI provisioning, and secret distribution,

  • application-specific misuse of opaque payloads or attachments.

Security Assumptions

The security properties described here rely on the following assumptions:

  • secure transports correctly validate peer credentials and protect link confidentiality/integrity,

  • implementations correctly validate message syntax, lengths, and extension chains,

  • local policy engines are configured with the intended trust anchors and authorization rules,

  • secret management and credential provisioning are handled securely outside the base protocol.

Timestamps help ordering and conflict resolution, but they are not by themselves a cryptographic anti-replay mechanism for arbitrary application data.