Skip to main content

Sovereign deployment threat model

Status and scope

This document defines the v1 threat and responsibility baseline for Fairway's Sovereign Deployment Ready reference work. It is an engineering security input, not a certification, legal conclusion, system authorization, Security Target, or claim that the current Fairway release implements every control below.

The baseline covers three bounded deployment shapes:

ProfileBoundaryExpected connectivity
sovereign-offlineOne customer-controlled host or isolated local network; local SQLite remains the execution store.No outbound DNS, HTTP, telemetry, update, remote provider, remote identity, or remote asset dependency during install and operation. Explicit loopback and local sockets only.
sovereign-connectedCustomer-controlled network with explicit internal services and allowlisted update/evidence transfer paths.Deny by default. Every identity, notifier, provider, registry, package, and update dependency is named and explicitly permitted by customer policy.
restricted-sharedShared-team Fairway service and store inside a customer-controlled restricted-data boundary.Authenticated, encrypted internal paths only. Shared writes require verified identity, scoped authorization, audit, conflict control, and separately reviewed deployment support.

The evaluated scope must name the Fairway version, source revision, binary and package digests, profile ID/version/digest, configuration identity, store mode, host or orchestration baseline, enabled adapters, trust roots, data boundary, and evidence-package digest. A result for one scope does not transfer to another version, configuration, deployment, customer, or jurisdiction.

Assets

  • Fairway binary, release bundle, checksums, signatures, SBOM, VEX, and build provenance;
  • project configuration, assurance profiles, policy versions, and deployment baseline;
  • Fairway task, decision, evidence-reference, review, checkpoint, wait, session, notification, audit, and state-history records;
  • assurance package manifests, normalized evidence indexes, readiness reports, exception links, and external-assessment references;
  • customer identity, authorization, network, key, backup, restore, retention, and trusted-time configuration;
  • offline trust roots, update media, rollback targets, security advisories, and disposal records.

Fairway does not make raw prompts, private transcripts, provider tool bodies, credentials, secrets, or arbitrary artifact contents part of the assurance record. External source-control, CI, package, identity, signing, backup, and deployment systems remain authoritative for facts they create.

Trust boundaries

BoundaryTrusted only forMust not imply
Human operatorExplicit local or external action under a named authorizationThat Fairway grants deployment, release, credential-use, or live-mutation authority
ReviewerAttributable verdict in a configured review domainProduct certification, customer risk acceptance, or self-review
DashboardRead-oriented display of Fairway stateWrite, send, approval, merge, deploy, release, or live authority
Provider or utility adapterBounded lifecycle, notification, or evidence metadataProvenance, approval, unrestricted command execution, or secret custody
Identity proxy or API tokenIdentity only after configured cryptographic or origin proofImplicit role, project, review, or consequential-action authority
Source/build/release systemsSource revision, build, artifact, and release facts they produceTrustworthiness beyond verified provenance and the evaluated builder boundary
Customer platformNetwork, host, identity, encryption, key custody, retention, and backup controlsFairway product implementation or external certification
External assessor or certification bodyFindings or certificates for its named scheme and exact evaluated scopeBroader versions, configurations, customers, jurisdictions, or unevaluated controls

Assumptions

  • The customer can enforce the selected host, network, identity, key, storage, backup, time, and physical-security boundary.
  • Offline installation begins with documented trust roots delivered separately from the bundle being verified.
  • Fairway can detect and report evidence gaps but cannot determine legal applicability or accept customer risk.
  • A compromised operating system, hypervisor, administrator, build platform, or signing root can invalidate evidence unless an independent control detects it.
  • Availability and denial-of-service protection depend partly on the customer platform and are not established by local evidence integrity alone.

Threats and required controls

ThreatAttack or failureRequired baseline controlsPrimary responsibilityResidual decision owner
T01 malicious-artifactA binary, plugin, profile, update, evidence package, or removable-media bundle is substituted or contains malicious content.SDR-PROVENANCE, SDR-ADAPTER-CONTAINMENT, SDR-VULNERABILITY-AND-SUPPORT, SDR-EVIDENCE-AND-CLAIMS; digest/signature verification before use; no automatic execution from evidence.Product and customerCustomer security owner
T02 build-compromiseBuild infrastructure emits a backdoored artifact or forged/incomplete provenance.SDR-PROVENANCE; trusted builder policy; source-to-subject verification; independent release review; reproducibility claims only when demonstrated.Product and external assessorProduct security owner
T03 operator-impersonationAn attacker acts as an operator through a stolen token, unverified proxy header, stale session, or identity fallback.SDR-IDENTITY-AND-AUDIT, SDR-CRYPTOGRAPHY-AND-KEYS; verified identity; scoped roles; expiry/revocation; attribution.SharedCustomer identity owner
T04 reviewer-impersonationA writer spoofs reviewer identity, self-reviews, replays a verdict, or applies a verdict outside its domain or commit.SDR-IDENTITY-AND-AUDIT, SDR-EVIDENCE-AND-CLAIMS; authenticated reviewer binding; review scope; anti-replay/idempotency; reviewed source identity.Product and customerReview-policy owner
T05 insider-misuseAn authorized user alters state, exports sensitive metadata, weakens policy, or performs a consequential action beyond need.SDR-IDENTITY-AND-AUDIT, SDR-DATA-MINIMIZATION; least privilege; separation of duties; explicit promotion boundaries; retention and review.SharedCustomer risk owner
T06 evidence-tamperRecords or packages are deleted, inserted, reordered, rewritten, backdated, or mixed across projects/scopes.SDR-IDENTITY-AND-AUDIT, SDR-EVIDENCE-AND-CLAIMS; append-only facts where possible; audit binding; fixed manifests; source-project and task checks; trusted time boundary.Product and customerEvidence owner
T07 rollbackAn old binary, policy, DB, trust root, or evidence snapshot is restored and presented as current.SDR-PROVENANCE, SDR-IDENTITY-AND-AUDIT, SDR-RECOVERY, SDR-VULNERABILITY-AND-SUPPORT; version/readback; externally anchored audit where required; approved rollback target; post-restore verification.SharedCustomer operations owner
T08 backup-lossBackups are absent, unreadable, exposed, use lost keys, or cannot restore the expected state.SDR-CRYPTOGRAPHY-AND-KEYS, SDR-RECOVERY; encrypted/versioned backups; restore drills; key recovery; retention; disposal proof.CustomerCustomer operations owner
T09 adapter-escapeA provider, notifier, utility, tracker, browser asset, or command adapter bypasses the network/data boundary or gains hidden authority.SDR-NETWORK-ISOLATION, SDR-ADAPTER-CONTAINMENT, SDR-DATA-MINIMIZATION; disabled-by-default remote adapters; allowlists; capability preflight; metadata-only records; no dashboard send authority.Product and customerCustomer platform owner
T10 data-boundary-failureSecrets, private paths, prompts, transcripts, raw tool bodies, evidence contents, or restricted metadata leave the policy-permitted boundary.SDR-NETWORK-ISOLATION, SDR-IDENTITY-AND-AUDIT, SDR-DATA-MINIMIZATION, SDR-EVIDENCE-AND-CLAIMS; data inventory; redaction; local roots; export allowlist; egress deny; negative tests.SharedCustomer data owner
T11 key-compromiseA signing, encryption, token, or backup key is exposed, substituted, unavailable, or not customer controlled.SDR-CRYPTOGRAPHY-AND-KEYS; external secret store; key identity and purpose; rotation/revocation/recovery; pinned verification roots.Customer and productCustomer key owner
T12 network-escapeDNS, redirects, proxy environment variables, remote assets, update checks, or adapters create unexpected outbound traffic.SDR-NETWORK-ISOLATION, SDR-ADAPTER-CONTAINMENT; deny-by-default network policy and egress-denied rehearsal.SharedCustomer network owner

Control baseline

ControlObjectiveResponsibilityMinimum evidence before a bounded readiness claimDelivery task
SDR-BOUNDARYBind every result to exact product, deployment, policy, customer boundary, profile, and evidence versions.SharedConfiguration identity, profile digest, source/release identity, scope statement, customer applicability decision.FW-341, FW-349, FW-354
SDR-NETWORK-ISOLATIONFail closed on undeclared DNS, HTTP, telemetry, update, identity, asset, notifier, and provider dependencies.SharedCapability inventory, egress-deny tests, redirect/proxy/DNS negative tests, disconnected rehearsal.FW-342
SDR-PROVENANCEVerify source, builder, artifact, release, SBOM/VEX, signature, and provenance subjects offline.ProductSigned release bundle, checksums, provenance, builder policy, SBOM, VEX, verification log.FW-343, FW-347
SDR-IDENTITY-AND-AUDITBind shared actions and review judgments to verified actors and detect audit tampering without storing private content.SharedIdentity proof, role policy, negative auth tests, self-review/replay tests, tamper verification, retention, revocation/expiry evidence.FW-344, FW-346
SDR-CRYPTOGRAPHY-AND-KEYSName every encryption/signing boundary, module, key owner, trust root, and recovery path.SharedCrypto inventory, module/version/config reference, customer key custody, rotation/recovery rehearsal, explicit FIPS non-claim unless externally supported.FW-345
SDR-RECOVERYRestore the exact supported state and safely roll back software, policy, data, and trust roots.CustomerBackup manifest, restore/readback proof, rollback target, compatibility result, key-loss procedure, cleanup proof.FW-348
SDR-ADAPTER-CONTAINMENTKeep providers, notifiers, trackers, utilities, and remote assets disabled or explicitly bounded.Product and customerAdapter inventory, allowlist/disable proof, network and authority negative tests, notification evidence.FW-342, FW-348
SDR-DATA-MINIMIZATIONKeep restricted content and identifiers inside the policy-permitted boundary and out of routine Fairway records/exports.SharedData inventory, redaction tests, artifact-root policy, export review, egress-deny evidence, retention/disposal policy.FW-346, FW-349
SDR-VULNERABILITY-AND-SUPPORTDeliver advisories, affected versions, mitigations, fixes, VEX, and offline updates through a supported lifecycle.ProductVulnerability policy, advisory fixtures, LTS/EOL policy, synthetic offline patch rehearsal.FW-353
SDR-OFFLINE-REHEARSALProve bounded install, operation, verification, recovery, rollback, and cleanup with outbound network denied.SharedNon-authoring disconnected rehearsal packet, exact timing/readback, failure and cleanup evidence.FW-350
SDR-INDEPENDENT-ASSESSMENTObtain a qualified security assessment for the exact scope.External assessor and customerAssessment report reference, findings/retest status, assessor identity, exact criteria, and explicit limitations.FW-352
SDR-EVIDENCE-AND-CLAIMSProduce deterministic packages and reject claims beyond recorded evidence and external authority.ProductSigned assurance package, offline verification, responsibility/gap matrix, claim manifest, public wording review.FW-349, FW-354, FW-355 through FW-361

No control is satisfied merely because it appears in this table. Readiness is computed from the selected profile and current evidence; customer and external assessment responsibilities remain gaps until their named owners provide proof.

Framework and evaluation-input crosswalk

This crosswalk is navigation for assessors. It is incomplete and does not claim equivalence, control inheritance, conformity, an Evaluation Assurance Level, or that one framework satisfies another.

Baseline areaNIST SSDF 1.1NIST SP 800-171 Rev. 3 familiesEU CRA 2024/2847 technical-documentation inputProduct-evaluation input
Scope, policy, rolesPO.1, PO.2Planning; Access Control; Identification and AuthenticationProduct description, intended purpose, versions, cybersecurity risk assessmentTarget of Evaluation boundary; assumptions; organizational security policies; security objectives
Secure environment and adaptersPO.5, PS.1Configuration Management; System and Communications Protection; System and Services AcquisitionArchitecture/design and solutions used for applicable essential requirementsOperational environment; attack surface; functional requirements and configuration assumptions
Source, build, release integrityPS.2, PS.3Supply Chain Risk Management; System and Services Acquisition; System and Information IntegrityDesign/development information, third-party components, tests, software bill of materials when applicableDevelopment, delivery, configuration-management, and life-cycle evidence
Vulnerability and supportRV.1, RV.2, RV.3Risk Assessment; Incident Response; System and Information IntegrityVulnerability-handling processes, support period, test reportsFlaw-remediation and vulnerability-analysis evidence
Audit, evidence, and assessmentPO.4, PW.7, PW.8Audit and Accountability; Security Assessment and MonitoringRisk assessment, test reports, applied specifications, technical documentation updatesSecurity problem definition, Security Target inputs, assurance evidence, evaluator findings
Recovery and continuityPO.5Media Protection; System and Information IntegrityProduct maintenance/support evidence where applicableOperational guidance, secure state, recovery assumptions, residual risks

Authoritative references:

Common Criteria terms are used only to structure future evaluation inputs: an exact Target of Evaluation, security problem definition, assumptions, threats, organizational security policies, objectives, functional/assurance requirement selection, and evidence. This baseline is not a Protection Profile, Security Target, CEM evaluation, EAL result, or certificate.

Validation and change control

The threat model and baseline are versioned together. A new threat, deployment mode, authority path, data class, adapter type, store, identity source, crypto module, public claim, or external framework version requires a profile diff, threat review, control/evidence update, and independent approval. Historical packages retain their original profile and evidence digests.