Report issues against the latest supported release on main.

Report security issues privately via GitHub Security Advisories:

https://github.com/fraware/SCOPE/security/advisories/new

Do not open public issues for undisclosed vulnerabilities. Include reproduction steps, affected version / commit, and impact. Maintainers will acknowledge and coordinate a fix before any public disclosure.

Security reviews for this release must cite threat_model_version = scope-threat-model-v2.2-as08 (or a later ADR that supersedes it).

SCOPE records human review authority and bounded grant use. It builds on schema validation, canonical hashing, hash-chained ledger events, explicit expiration checks, and fail-closed behavior for unknown scopes, invalid roles, forbidden queue transitions, incomplete independence blocks, and unauthorized pilot modes.

Cryptography and signing

• Ed25519 signatures on decisions and grants when production signing is enabled
• Signing assurance levels (SAL0–SAL4) with KMS reference adapter (--signing-provider kms)
• Combined scope_trust_root_hash ties policy and reviewer registry integrity into PCS export provenance
• Frozen packets and deep-frozen grants reject in-place mutation after finalize

Identity

• Identity assurance levels (IAL0–IAL4); production mode rejects IAL0 unless SCOPE_ALLOW_DEV_IAL0
• Pluggable OIDC and SAML providers (scope/identity_providers.py)
• SCIM/LDAP RBAC sync via scope rbac sync
• Live IdP certification remains external (see Definition of Done)

Ledger and delivery

• Local hash-chained JSONL / SQLite / Postgres ledger with WORM and verified remote sink options
• Delivery modes: best_effort, at_least_once, fail_closed
• Live WORM Object Lock and remote ledger authority remain external

AKTA review contract

• Frozen at scope-akta-review-v1.0; consumers branch on summary.status

Workflow

• Tenant-isolated review queues via authenticated principal binding
• Pilot review_mode enforcement: shadow / read-only modes cannot authorize
• Append-only outcome links; decisions are never rewritten after outcomes

Known limits

• SAML verification requires external sidecar or pre-verified assertions in reference adapter
• Email notifications require institutional SMTP wiring
• Reviewer judgment and physical lab safety remain out of scope
• Quality metrics are process/calibration aids, not scientific ground truth

python scripts/run_institutional_offline_gates.py

See docs/threat_model.md, docs/trusted_boundary.md, docs/compatibility_matrix.md, docs/limitations.md, docs/definition_of_done.md, docs/adr/, and docs/runbooks/.