Skip to main content

Threat model

This model covers the reviewed V2 backend, current agent foundation, and documented target endpoint pipeline. It assumes an attacker may control an endpoint or submit hostile external feed content; possession of infrastructure credentials is treated as a separate compromise. Development defaults do not establish an internet-safe production perimeter. It synthesizes the existing backend threat models and source rather than certifying a particular deployment.

Assets and authority

Primary assets are telemetry integrity, tenant isolation, operator decisions, endpoint private keys, the signing authority, retained analytical evidence, ruleset integrity, and vulnerability intelligence provenance. The most consequential writers are the control process, authorized broker producers, privileged database accounts, and endpoint configuration/package administrators.

Threats and controls

Threat / affected boundaryCurrent mitigationResidual risk and status
Forged agent at gatewayVerified client chain; identity from certificate; registry admissionStolen authorized key can impersonate its agent; implemented authentication, no endpoint attestation
Malicious enrolled endpointTyped validation, size/age/rate bounds, stamped originCan lie within its permitted telemetry; mTLS cannot establish observation truth
Stolen enrollment tokenNo bootstrap-token service existsToken expiry/use-count/replay controls belong to planned automated enrollment
MITM or backend impersonationBackend mTLS; validated trust bundle in agent configurationInitial trust provisioning must be authenticated; live agent transport remains planned
Stolen endpoint key / cloned stateOwner-only state and key files; explicit agent revocationRoot or the account can copy files; hardware non-exportability and clone handling are not implemented
Telemetry replay / duplicate publicationStable identities, replay-aware consumers, bounded event ageRequest retries can duplicate broker records; no global exactly-once claim
Event loss during crash/outageBroker-before-ACK; store-before-offset commitBroker retention/cluster loss remain limits; endpoint spool is planned
Unauthorized control requestCertificate-bound session, per-request policy, tenant checksPolicy/CA compromise grants authority; sessions remain local to a process
Query access survives policy changeCertificate-derived tenant scope is enforcedQuery is not yet integrated with control policy; certificate scope must be managed separately
Forged admission/ruleset/advisory broker recordExplicit ownership and topology; TLS/SCRAM supportProducer ACL provisioning is incomplete in supplied deployment; high-impact infrastructure trust boundary
Database compromiseSeparate transactional and analytical workloadsPrivileged store access can expose/alter data; complete production role/backup hardening is deployment work
Malicious configuration or privileged moduleOwnership checks; atomic local config activation; dependency boundariesIn-process modules share privileges; signed central policy and a helper are planned
Tampered update / dependencyLocked/pinned dependencies and existing CI gatesNo complete signed native agent release/update channel yet
Hostile advisory feedBounded fetch/parse, schema translation, platform-written provenanceProvenance identifies bytes, not truth; compromised upstream can supply misleading intelligence
Resource exhaustionGateway capacity/rate bounds; bounded windows and session storesSustained authorized input can consume allowed capacity; protection requires monitoring and sizing

Operational priorities

Protect CA and broker producer credentials first because they can create identities or authoritative records, not merely read data. Isolate development services from untrusted networks. Review query access revocation separately from control policy. Preserve incident and alert trails in backups. Observe failure and saturation signals before tuning their ceilings upward.

Future boundaries requiring fresh analysis

Enrollment token redemption, spool recovery, signed policy, native packaging, verified upgrades, privileged collection, and response execution each add an attack surface. Their plan must not be read as evidence that the required mitigation already exists. Revisit this model when those components enter the composition root.

Source evidence

Reviewed against the source baseline. Seagull-backend-v2/notes/Seagull-backend-v2-threat-model.md (local planning evidence) · Seagull-backend-v2/notes/Seagull-backend-v2-advisories-threat-model.md (local planning evidence) · Seagull-backend-v2/internal/control/guard.go · Seagull-backend-v2/internal/broker/security.go · Seagull-agent-v2/README.md.