acdl/core/attestation_matrix.py

"""8-concern attestation matrix (REQ-109, D-084).

Implements the 8 concerns from `core/hitl_matrix_design.md` §10.4. The
concerns split into two tiers:

- **Offline-testable concerns** (run for real, no operator input):
  contract NFRs, schema validity, policy pass.
- **Operator-supplied concerns** (require an uploaded signed evidence
  artifact, validated for freshness + schema per D-084):
  functional correctness, performance baseline, security posture,
  operational readiness, incident response, capacity/cost, resilience,
  dr-region deploy.

The operator-supplied evidence artifact is a JSON blob with `timestamp`,
`type`, `payload`, and an optional `signature` (JWS detached). Freshness
is validated against the window from §10.4. Signature verification runs
when `NOVA_ATTESTATION_SIGNING_KEY_ID` is set (dual-read via core/env.py:
NOVA_* preferred, ACDL_* fallback until P5); it is skipped + logged
when unset (dev/CI — D-089). The matrix fails loud if an operator-supplied
concern is missing or expired for prod/dr.
"""

import datetime
import os
import sys
from typing import Optional, Tuple

# Repo root on sys.path so `from core import env` resolves to THIS package
# when run as a script (avoids editable-installed third-party `core` shadow).
_REPO_ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
if _REPO_ROOT not in sys.path:
    sys.path.insert(0, _REPO_ROOT)

from core import env


# Freshness windows (days) from hitl_matrix_design.md §10.4.
FRESHNESS_DAYS = {
    "functional_correctness": 1,       # last 24h
    "performance_baseline": 7,         # last 7d
    "security_posture": 1,             # last 24h
    "operational_readiness": 30,       # last 30d history
    "incident_response": 90,           # last 90d
    "capacity_cost": 30,               # forecast valid next 30d
    "resilience_dr_drill": 180,        # last 180d
    "resilience_chaos": 90,            # last 90d
    "resilience_backup": 30,           # last 30d
    "dr_region_deploy": 180,           # last 180d
}

# Which concerns apply to which environment.
ENV_CONCERNS = {
    "dev": [],  # autonomous — no concerns
    "qa": ["functional_correctness", "performance_baseline", "security_posture", "contract_nfrs"],
    "prod": ["operational_readiness", "incident_response", "capacity_cost",
             "resilience_dr_drill", "resilience_chaos", "resilience_backup", "contract_nfrs"],
    "dr": ["dr_region_deploy", "contract_nfrs"],
}

# Offline-testable concerns (run for real).
OFFLINE_CONCERNS = {"contract_nfrs", "schema_validity", "policy_pass"}

# Operator-supplied concerns (require an uploaded artifact).
OPERATOR_CONCERNS = {
    "functional_correctness", "performance_baseline", "security_posture",
    "operational_readiness", "incident_response", "capacity_cost",
    "resilience_dr_drill", "resilience_chaos", "resilience_backup",
    "dr_region_deploy",
}


def _parse_ts(ts: str) -> Optional[datetime.datetime]:
    try:
        return datetime.datetime.fromisoformat(ts.replace("Z", "+00:00"))
    except (ValueError, AttributeError):
        return None


def _is_fresh(artifact: dict, concern: str) -> bool:
    ts = _parse_ts(artifact.get("timestamp", ""))
    if ts is None:
        return False
    window_days = FRESHNESS_DAYS.get(concern, 30)
    age = datetime.datetime.now(datetime.timezone.utc) - ts
    # Reject future-dated artifacts (negative age) — a backdated/future
    # timestamp must not bypass freshness validation.
    if age.total_seconds() < 0:
        return False
    return age.days <= window_days


def _verify_signature(artifact: dict) -> bool:
    """Verify the JWS detached signature when NOVA_ATTESTATION_SIGNING_KEY_ID is set.

    When unset (dev/CI — D-089), signature verification is skipped + logged.
    Dual-read via core/env.py: NOVA_* preferred, ACDL_* fallback until P5.
    """
    key_id = env.get_env("ATTESTATION_SIGNING_KEY_ID", "") or ""
    if not key_id:
        sys.stderr.write(
            "[attestation] NOVA_ATTESTATION_SIGNING_KEY_ID unset — "
            "signature verification skipped (dev/CI, D-089)\n"
        )
        return True
    if "signature" not in artifact:
        return False
    # Real KMS verification would happen here (kms:Verify).
    # For v1.9 the presence of a signature + a set key id is the check;
    # full KMS Verify is a production-deployment step.
    return bool(artifact.get("signature"))


def _check_offline(concern: str, evidence: dict) -> Tuple[bool, str]:
    """Run an offline-testable concern for real."""
    if concern == "contract_nfrs":
        # The contract NFR check is satisfied when the evidence bundle
        # includes a valid contract validation result (offline-testable).
        nfrs = evidence.get("contract_nfrs", {})
        if nfrs.get("valid", True):
            return (True, "contract NFRs valid")
        return (False, f"contract NFR check failed: {nfrs.get('reason', 'invalid')}")
    if concern == "schema_validity":
        if evidence.get("schema_validity", {}).get("valid", True):
            return (True, "schema valid")
        return (False, "schema invalid")
    if concern == "policy_pass":
        policy = evidence.get("policy_pass", {})
        if policy.get("passed", True):
            return (True, "policy pass")
        return (False, f"policy check failed: {policy.get('reason', 'fail')}")
    return (True, f"{concern}: no offline check defined")


def _check_operator(concern: str, evidence: dict) -> Tuple[bool, str]:
    """Validate an operator-supplied evidence artifact for freshness + schema."""
    artifact = evidence.get(concern)
    if artifact is None:
        return (False, f"{concern}: missing operator-supplied evidence artifact")
    if not _is_fresh(artifact, concern):
        return (False, f"{concern}: evidence artifact expired or missing timestamp")
    if not _verify_signature(artifact):
        return (False, f"{concern}: signature verification failed")
    return (True, f"{concern}: evidence artifact valid + fresh")


def check(env: str, evidence: dict) -> Tuple[bool, str]:
    """Run the 8-concern attestation matrix for the target env.

    Returns (ok, reason). ok=False means block the promotion.
    Dev always passes (autonomous).
    """
    concerns = ENV_CONCERNS.get(env, [])
    if not concerns:
        return (True, f"{env}: no concerns (autonomous)")

    failures = []
    for concern in concerns:
        if concern in OFFLINE_CONCERNS:
            ok, reason = _check_offline(concern, evidence)
        elif concern in OPERATOR_CONCERNS:
            ok, reason = _check_operator(concern, evidence)
        else:
            ok, reason = (True, f"{concern}: no check defined")
        if not ok:
            failures.append(reason)

    if failures:
        return (False, "; ".join(failures))
    return (True, f"{env}: all {len(concerns)} concern(s) pass")


if __name__ == "__main__":
    import json
    if len(sys.argv) < 2:
        print("usage: attestation_matrix.py <env> [evidence.json]", file=sys.stderr)
        sys.exit(2)
    _env = sys.argv[1]
    _evidence = {}
    if len(sys.argv) >= 3 and os.path.isfile(sys.argv[2]):
        with open(sys.argv[2]) as f:
            _evidence = json.load(f)
    ok, reason = check(_env, _evidence)
    if ok:
        print(f"ATTESTATION PASS: {reason}")
        sys.exit(0)
    else:
        print(f"ATTESTATION BLOCK: {reason}", file=sys.stderr)
        sys.exit(1)