Commit Graph

3 Commits

Author SHA1 Message Date
shankar0123 5e2accbf5f auth-bundle-2 Phase 12: extend auth-threat-model.md with Bundle 2 sections (OIDC + sessions + back-channel logout + OIDC first-admin + break-glass + 8 Bundle 2 threat sub-sections)
Closes Phase 12 of cowork/auth-bundle-2-prompt.md. The single
canonical operator-facing threat model (one doc per topic per the
docs convention) now covers both Bundle 1 (RBAC) AND Bundle 2 (OIDC
+ sessions + back-channel logout + OIDC first-admin + break-glass)
in one place.

File: docs/operator/auth-threat-model.md (MODIFIED, +485 LOC)

Conventions held
================

* The Bundle 1 sections ("Threat actors", "Defenses Bundle 1
  ships", "Threats Bundle 1 does NOT close", "Compliance mapping",
  "Operator-facing checks", "Cross-references") stay structurally
  intact. Bundle 2 EXTENDS them; nothing is rewritten in place.
* `Last reviewed:` header bumped 2026-05-09 → 2026-05-10.
* Per the prompt's explicit instruction: "do NOT create a separate
  auth-threat-model-bundle-2.md companion." This commit is a
  single-file extension.

Changes
=======

Intro paragraph rewritten:
* From "Bundle 1 lands... Bundle 2 will be updated" to "Bundle 1
  AND Bundle 2 land." Sets the reader's expectation that this is
  the post-Bundle-2 doc.

Threat actors section (4 new actors appended):
* OIDC-federated end user (token-forgery / session-hijacking /
  group-claim-manipulation surface).
* Stolen session cookie holder (XSS / network MITM / pasted-token).
* Compromised IdP (rogue token issuance; mitigations bounded to
  audit trail + group-mapping configuration).
* Break-glass-password holder (Phase 7.5 path bypasses OIDC + group
  layer entirely; default-OFF is the load-bearing mitigation).

NEW: Defenses Bundle 2 ships (5 sub-sections):
* OIDC token validation (Phase 3) — alg allow-list, IdP-downgrade
  defense, exact iss match, aud + azp checks, at_hash
  REQUIRED-when-access_token-present (Phase 3 tightening of OIDC
  core's MAY → MUST), single-use state + nonce, PKCE-S256 mandatory,
  iat window, JWKS rotation handling, JWKS-fetch-fail closed,
  encrypted client_secret at rest.
* Session minting + cookies (Phases 4 + 6) — length-prefixed HMAC
  defeating concatenation collision, HttpOnly + Secure + SameSite
  cookie hardening, idle + absolute timeouts, CSRF defense via
  double-submit-cookie + hashed-token-on-row, optional IP/UA bind,
  signing-key rotation primitive with retention window, fail-fatal
  EnsureInitialSigningKey at boot, pre-login vs post-login cookie
  discrimination.
* Back-channel logout (Phase 5) — OpenID Connect Back-Channel
  Logout 1.0 (NOT RFC 8414), required-claim pinning, jti-based
  replay defense, alg allow-list applies, Cache-Control: no-store.
* OIDC first-admin bootstrap (Phase 7) — coexists with Bundle 1's
  env-var-token bootstrap, group-scoped, one-shot per tenant via
  admin-existence probe, explicit OIDC provider gate, audit row on
  every grant.
* Break-glass admin (Phase 7.5) — default-OFF, surface-invisibility
  via 404-not-403, Argon2id with OWASP 2024 params, lockout state
  machine, constant-time across all failure paths via verifyDummy,
  WARN log at boot when ENABLED=true, 5/min rate limit on the
  public login endpoint.

NEW: Bundle 2 threat catalogue (8 sub-sections, one per
prompt-enumerated threat axis):

1. OIDC token forgery vectors and mitigations (9-row table covering
   alg confusion, audience injection, issuer mismatch, nonce replay,
   state replay, at_hash substitution, iat window manipulation,
   JWKS rotation mid-login, JWKS-fetch failure during a key
   rotation).
2. Session hijacking vectors and mitigations (7-row table covering
   XSS cookie theft, network MITM, CSRF, concatenation-collision
   forgery, stolen-cookie replay, cross-tab interference, sign-out
   race).
3. IdP compromise scenarios (operator monitors IdP audit logs,
   operator can rotate group-role mappings without redeploying,
   audit trail records source provider, provider-delete returns
   409 with active sessions).
4. Back-channel logout failure modes (6-row table covering IdP
   unreachable, invalid signature, replay via jti, alg confusion,
   missing events claim, present-nonce-claim).
5. Group-claim manipulation (4-row table covering operator
   misconfigured mapping, misconfigured groups_claim_path, IdP
   renames a group, IdP user maintainer adds user to unintended
   group).
6. Bootstrap phase risks post-Bundle-2 (4-row table covering
   CERTCTL_BOOTSTRAP_TOKEN leak, CERTCTL_BOOTSTRAP_ADMIN_GROUPS
   misconfigured to a wide group, both bootstrap strategies
   simultaneously, multi-IdP without explicit provider gate).
7. Break-glass risks (7-row table covering phished password,
   online brute-force, offline brute-force on DB compromise,
   operator forgets to disable, side-channel timing on
   wrong-vs-no-credential-vs-locked, surface fingerprinting,
   reserved-actor mutation).
8. Token-leak hygiene (the explicit grep policy with three
   per-package logging_test.go pointers + the audit_redact.go
   defense-in-depth note).

Threats Bundle 1 does NOT close section relabeled:
* Section header now reads "Threats Bundle 1 does NOT close
  (Bundle 2 closure status)" with each item carrying  / ⚠️ /
  "still deferred" markers.
* Items 1, 2, 3, 8 marked  closed by Bundle 2.
* Items 4, 5, 7, 9 marked still-deferred with v3 / follow-on
  pointers.
* Item 6 (rate limiting on bootstrap) marked acceptable; Bundle 2
  adds the same rate-limit primitive to /auth/breakglass/login.

NEW: Threats Bundle 2 does NOT close section listing the 8 v3 /
future-work items:
* WebAuthn / FIDO2 second factor (Decision 12).
* Time-bound role grants / JIT elevation.
* SAML federation (operators broker through Keycloak).
* Multi-tenant data isolation activation (gated to managed-service
  hosting work).
* HSM / FIPS-validated signing key for sessions.
* OIDC RP-initiated logout (Bundle 2 implements only back-channel).
* GUI E2E via Playwright.
* Per-IdP runbook external-tester sign-off (encouraged, NOT a merge
  gate post-2026-05-10 policy change).

Operator-facing checks section extended:
* 6 new SQL-shaped checks for Bundle 2 (provider count drift,
  per-actor session count, unmapped-groups audit-row spike,
  break-glass usage outside incidents, OIDC first-admin one-row-per-
  tenant invariant, retired-signing-key GC liveness).

Cross-references section split into Bundle 1 anchors + Bundle 2
anchors:
* Bundle 2 anchors enumerate every load-bearing file: 6
  internal/auth/ packages, 5 migrations, 3 ci-guards.

Compliance mapping section UNCHANGED:
* Phase 15 (standards-and-RFC-implementation table) is the proper
  home for the RFC + CWE evidence the Bundle 2 surface adds.
  Re-introducing framework-mapping prose at the threat-model layer
  would regress the operator's 2026-05-05 retired-compliance-docs
  decision, which is explicitly forbidden by the Phase 15 prompt.

Verification
============

* `> Last reviewed: 2026-05-10` — confirmed via head -3.
* All 8 prompt-mandated Bundle 2 threat sub-sections present —
  confirmed via grep `^### ` count (19 ### headers total: 6 Bundle
  1 + 5 Bundle 2 defenses + 8 Bundle 2 threats).
* All 39 prompt-listed threat-vector keywords present — confirmed
  via single-line grep counting 39 hits across the prompt's
  vocabulary.
* Internal markdown links resolve cleanly — confirmed via shell
  loop iterating each `]( ...)` reference and checking `[ -e "$path" ]`.
* No backend / Go-test impact — pure docs commit.
* `make verify` gate unchanged.
2026-05-10 16:11:08 +00:00
shankar0123 5313cd8492 auth-bundle-1 Phase 13 follow-up: em-dash sweep + broken-link fix
Self-audit on e7a94b6 flagged the prompt's 'zero em dashes'
discipline rule. The four new Phase 13 docs and the v2.1.0
CHANGELOG section had 97 em-dash hits between them; this commit
sweeps them all to ASCII hyphens.

Counts before -> after:
  docs/operator/rbac.md                  28 -> 0
  docs/operator/auth-threat-model.md     36 -> 0
  docs/migration/api-keys-to-rbac.md     16 -> 0
  docs/operator/security.md               8 -> 0
  docs/reference/profiles.md              3 -> 0
  CHANGELOG.md                            6 -> 0

Mechanical: ' - ' (spaced em dash) and bare em-dash both replaced
with spaced ASCII hyphen, then double-spaces collapsed. Markdown
list bullets ('^- ', '^  - ', '^    - ') verified intact across
all six files. Internal-link sweep also re-run.

Also fixes a pre-existing broken link the audit caught:
  docs/operator/security.md:70 referenced
  '../internal/crypto/encryption.go' which is a 1-level-up jump
  from docs/operator/, not the 2-level-up jump it actually needs
  ('../../internal/crypto/encryption.go'). Pre-Bundle-1 link rot;
  fixed in lockstep so the merge gate's docs validation passes
  cleanly.

Final state across the Phase-13 docs + CHANGELOG:
  - 0 em dashes
  - 0 broken internal links
  - Last-reviewed: 2026-05-09 header on every new doc

Bundle 1 documentation is now ready for the operator-side merge
gate review.
2026-05-10 00:15:30 +00:00
shankar0123 e7a94b6080 auth-bundle-1 Phase 13: docs (rbac.md + threat model + migration guide + security.md update)
Closes the last Phase before the Bundle 1 Exit gate. Operators
now have authoritative reference + threat model + migration guide
covering every behavior change Bundles 0-12 introduced.

# New docs

* docs/operator/rbac.md (340 lines) — operator how-to:
  - Mental model (actors / roles / permissions / scopes)
  - 7 default roles seeded by migration 000029 + the 5
    admin-only fine-grained perms seeded by 000030
  - Permission catalogue table by namespace
  - Scope semantics (global beats specific) + the Bundle-2
    deferral on scope_id FK enforcement
  - Granting / revoking access from GUI + CLI + HTTP API + MCP
  - The auditor pattern (audit-only, no resource read)
  - Day-0 bootstrap flow (CERTCTL_BOOTSTRAP_TOKEN → curl →
    HTTP 410 thereafter)
  - Demo-mode (CERTCTL_AUTH_TYPE=none) caveat for production

* docs/operator/auth-threat-model.md (180 lines) — what the
  controls defend against:
  - 5 threat actors (external, wrong-role, compromised key,
    insider operator, compromised auditor)
  - Per-defense walk-through (API-key auth, RBAC, bootstrap,
    approval workflow + Phase 9 closure, audit trail,
    protocol-endpoint allowlist)
  - 9 explicit deferrals (OIDC, sessions, local accounts,
    JIT elevation, MFA, etc.) — Bundle 2 / future scope
  - Compliance mapping (SOC 2 CC6.1/CC6.3, HIPAA §164.312(b),
    NIST SSDF PO.5.2, FedRAMP AU-9, PCI-DSS §10)
  - 5 operator-runnable sanity checks (e.g.,
    'SELECT FROM audit_events WHERE actor=system-bypass' MUST
    return 0 in production)

* docs/migration/api-keys-to-rbac.md (200 lines) — v2.0.x →
  v2.1.0 upgrade flow:
  - The SECURITY: AUDIT YOUR API KEYS callout
  - Migration list (000029-000033) + what each does
  - 4-mode scope-down flow (interactive / non-interactive
    JSON / --suggest / --suggest --apply)
  - What changes for code that called auth.IsAdmin
  - Helm-specific upgrade flow with example post-upgrade Job
  - Docker Compose upgrade flow + the 5 examples folders
    that ride demo mode unchanged
  - Verification queries + rollback flow

# Updated docs

* docs/operator/security.md — Last-reviewed bumped to
  2026-05-09; existing Authentication-surface section
  extended to call out the Bundle 1 RBAC primitive,
  day-0 bootstrap path, and approval-bypass closure with
  cross-references to the new docs.

* docs/reference/profiles.md — Last-reviewed header
  formatting fixed (added the > blockquote prefix used
  consistently across the docs tree).

# docs/README.md navigation

* Operator section gains 2 new rows (RBAC + auth-threat-model)
  and Approval-workflow row updated to mention Phase 9
  closure.
* Reference section gains the Profiles row.
* Migration section gains the api-keys-to-rbac row with the
  AUDIT YOUR API KEYS callout in the link description.

# CHANGELOG.md v2.1.0 section refreshed

The Phase 7 commit landed the SECURITY: AUDIT YOUR API KEYS
callout. This commit appends the missing Phase 9-12 highlights:

  - Approval-bypass closure (profile-edit gate + flip-flop
    loophole + ErrApproveBySameActor invariant)
  - GUI: Roles / API Keys / Auth Settings / Approvals queue
  - 12 new MCP RBAC tools
  - Coverage gates on internal/auth + internal/service/auth
  - Protocol-endpoint allowlist pinned at 3 layers

Trailing cross-reference block now points at all 4 new docs.

# Verifications

* Every internal link in the 4 new/modified docs validated by
  shell sweep (find broken links → 0 hits).
* Every new doc carries 'Last reviewed: 2026-05-09' header
  with the > blockquote prefix matching the docs-tree
  convention.
* go vet ./... clean.
* staticcheck across every Bundle-1-touched Go package clean.
* gofmt -l clean repo-wide.
* go test -short -count=1 green across internal/auth (incl.
  bootstrap), internal/api/handler, internal/api/router,
  internal/cli, internal/service (incl. auth),
  internal/domain/auth, internal/mcp, cmd/cli (cmd/server
  has 1 environmental failure on the sandbox virtiofs-tmp:
  TestPreflightSCEPRACertKey_KeyWorldReadable_Refuses depends
  on tmpfs file-mode semantics that virtiofs propagates
  differently — pre-existing, unrelated to Bundle 1).
* Frontend: 19 Vitest tests across src/pages/auth/ +
  AuditPage all pass; tsc --noEmit clean.
2026-05-10 00:10:15 +00:00