feat(use-cases): deterministisches source_regulation-Mapping + Primaerzweck [migration-approved]
CI / detect-changes (push) Successful in 8s
CI / branch-name (push) Has been skipped
CI / guardrail-integrity (push) Has been skipped
CI / secret-scan (push) Has been skipped
CI / dep-audit (push) Has been skipped
CI / sbom-scan (push) Has been skipped
CI / build-sha-integrity (push) Failing after 4s
CI / validate-canonical-controls (push) Successful in 12s
CI / loc-budget (push) Successful in 14s
CI / go-lint (push) Has been skipped
CI / python-lint (push) Has been skipped
CI / nodejs-lint (push) Has been skipped
CI / nodejs-build (push) Has been skipped
CI / test-go (push) Has been skipped
CI / iace-gt-coverage (push) Has been skipped
CI / test-python-backend (push) Successful in 31s
CI / test-python-document-crawler (push) Has been skipped
CI / test-python-dsms-gateway (push) Has been skipped

Use-Case-Zuordnung jetzt DETERMINISTISCH aus der Quell-Regulierung (statt
LLM/scope-category): control_parent_links.source_regulation (79% der 13.588
MCs) -> Keyword-Mapper -> ~30 Domaenen-Use-Cases. 117/117 Regulierungen
gemappt (dse 44 Leitlinien, code_security 10, network_security 9, ...).

- use_case_registry.py: 37 Use Cases (Doku + Security + Produkt/Sektor:
  cra/ai_act/mica/mdr/maschinen/batterie/ehds/dsa/dma/psd2/aml/lksg/...) +
  use_case_for_regulation() Keyword-Mapper (117 Regulierungen abgedeckt).
- migration 150: is_primary auf mc_use_case_mappings + neue mc_regulations
  (MC->source_regulation, n:m, is_primary) als feine Filter-Dimension.
- classify_mc_use_cases.py: source_regulation-getriebener Seed; Primaerzweck =
  dominante Regulierung, Mehrfachzwecke = weitere. PYTHONPATH-Bootstrap.
- 18 Registry-Tests gruen (Mapper-Abdeckung + Konsistenz-Invariante).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Benjamin Admin
2026-06-09 16:27:06 +02:00
parent a48e919caa
commit 6ca4dcde3e
4 changed files with 354 additions and 47 deletions
@@ -28,7 +28,9 @@ VERIFICATION_METHODS: tuple[str, ...] = (
"manual", # menschliche Attestierung
)
USE_CASE_GROUPS: tuple[str, ...] = ("document", "security", "cross_cutting")
USE_CASE_GROUPS: tuple[str, ...] = (
"document", "security", "cross_cutting", "product",
)
@dataclass(frozen=True)
@@ -136,9 +138,188 @@ _USE_CASES: tuple[UseCase, ...] = (
"hybrid"),
categories=("security", "governance", "operations"),
keyword_tokens=("tisax", "vda", "prototypenschutz")),
UseCase("kritis", "KRITIS / NIS2-Umsetzung", "security",
regulations=("KRITIS-Dachgesetz", "BSI-KritisV", "BSIG"),
verification_methods=("it_process", "network", "document")),
UseCase("dora", "DORA (Digital Operational Resilience)",
"cross_cutting", regulations=("DORA",),
verification_methods=("it_process", "document", "network")),
# ── Produkt-/Sektor-Use-Cases (je Quell-Regulierung) ────────────
UseCase("ai_act", "KI-Verordnung (AI Act)", "product",
regulations=("KI-Verordnung", "AI Act", "NIST AI RMF"),
verification_methods=("document", "it_process", "source_code")),
UseCase("mica", "Markets in Crypto-Assets (MiCA)", "product",
regulations=("MiCA",),
verification_methods=("document", "it_process")),
UseCase("mdr", "Medizinprodukte (MDR)", "product",
regulations=("MDR",),
verification_methods=("document", "source_code", "it_process")),
UseCase("maschinen", "Maschinenverordnung", "product",
regulations=("Maschinenverordnung",),
verification_methods=("document", "source_code", "it_process")),
UseCase("batterie", "Batterieverordnung", "product",
regulations=("Batterieverordnung", "Batteriegesetz"),
verification_methods=("document", "it_process")),
UseCase("ehds", "European Health Data Space", "product",
regulations=("EHDS",),
verification_methods=("document", "it_process", "source_code")),
UseCase("produktsicherheit", "Produktsicherheit (GPSR)", "product",
regulations=("Produktsicherheitsverordnung", "EU Blue Guide"),
verification_methods=("document", "it_process")),
UseCase("dsa", "Digital Services Act", "product",
regulations=("Digital Services Act",),
verification_methods=("document", "it_process")),
UseCase("dma", "Digital Markets Act", "product",
regulations=("Digital Markets Act",),
verification_methods=("document", "it_process")),
UseCase("data_governance", "Data Act / Data Governance Act", "product",
regulations=("Data Act", "Data Governance Act"),
verification_methods=("document", "it_process")),
UseCase("zahlungsdienste", "Zahlungsdienste (PSD2)", "product",
regulations=("Zahlungsdiensterichtlinie",),
verification_methods=("document", "it_process", "source_code")),
UseCase("geldwaesche", "Geldwäsche (AML/GwG)", "product",
regulations=("AML-Verordnung", "GwG"),
verification_methods=("document", "it_process")),
UseCase("lieferkette", "Lieferkettensorgfalt (LkSG)", "product",
regulations=("LkSG",),
verification_methods=("document", "it_process")),
UseCase("whistleblowing", "Hinweisgeberschutz (HinSchG)", "product",
regulations=("HinSchG",),
verification_methods=("document", "it_process")),
UseCase("barrierefreiheit", "Barrierefreiheit (EAA)", "product",
regulations=("European Accessibility Act",),
verification_methods=("document", "source_code")),
# ── Weitere Rechts-Use-Cases (Dokument) ─────────────────────────
UseCase("verbraucherschutz", "Verbraucherschutz", "document",
regulations=("Konsumentenschutzgesetz",
"Digitale-Inhalte-Richtlinie"),
verification_methods=("document",)),
UseCase("urheberrecht", "Urheberrecht", "document",
regulations=("UrhG", "DSM-Urheberrechtsrichtlinie"),
verification_methods=("document",)),
UseCase("wettbewerbsrecht", "Wettbewerbsrecht (UWG)", "document",
regulations=("UWG",), verification_methods=("document",)),
UseCase("gleichbehandlung", "Gleichbehandlung (AGG)", "document",
regulations=("AGG",),
verification_methods=("document", "it_process")),
UseCase("steuerrecht", "Steuerrecht", "document",
regulations=("Abgabenordnung", "BAO"),
verification_methods=("document", "it_process")),
UseCase("handelsrecht", "Handelsrecht", "document",
regulations=("HGB", "UGB", "ABGB"),
verification_methods=("document", "it_process")),
)
# Deterministischer Quell-Regulierung → Use-Case-Mapper.
# control_parent_links.source_regulation (117 distinct) → ~30 Domaenen.
# Reihenfolge = SPEZIFISCH zuerst (Substring, case-insensitive); der erste
# Treffer gewinnt. Die Datenschutz-Catch-alls (edpb/dsk/dsgvo) ganz hinten,
# damit spezifische Regeln (z.B. 'DSK OH Telemedien') zuerst greifen koennen.
_REGULATION_RULES: tuple[tuple[str, str], ...] = (
# Security / Code (Security-Team)
("owasp", "code_security"),
("nist sp 800-218", "code_security"),
("ssdf", "code_security"),
("nist sp 800-63", "code_security"),
("nistir 8259", "code_security"),
("cisa", "code_security"),
("nist sp 800-207", "network_security"),
("zero trust", "network_security"),
("nis2", "network_security"),
("nis-2", "network_security"),
("enisa", "network_security"),
("bsi-gesetz", "network_security"),
("bsig", "network_security"),
("cybersecurity act", "network_security"),
("kritis", "kritis"),
("nist cybersecurity framework", "isms"),
("nist sp 800-53", "isms"),
("digital operational resilience", "dora"),
("dora", "dora"),
# Produkt / Sektor
("cyber resilience act", "cra"),
("(cra)", "cra"),
("nist ai risk", "ai_act"),
("ki-verordnung", "ai_act"),
("ki-vo", "ai_act"),
("ai act", "ai_act"),
("oecd ki", "ai_act"),
("crypto-assets", "mica"),
("mica", "mica"),
("medizinprodukte", "mdr"),
("(mdr)", "mdr"),
("maschinenverordnung", "maschinen"),
("batterie", "batterie"),
("health data space", "ehds"),
("produktsicherheit", "produktsicherheit"),
("blue guide", "produktsicherheit"),
("digital services act", "dsa"),
("digital markets act", "dma"),
("data act", "data_governance"),
("data governance", "data_governance"),
("zahlungsdienste", "zahlungsdienste"),
("geldwaesche", "geldwaesche"),
("aml-verordnung", "geldwaesche"),
("lieferkettensorgfalt", "lieferkette"),
("lksg", "lieferkette"),
("hinweisgeberschutz", "whistleblowing"),
("hinschg", "whistleblowing"),
("accessibility act", "barrierefreiheit"),
# Website / Telemedien / Recht (User-Domaene)
("tdddg", "cookie_banner"),
("eprivacy", "cookie_banner"),
("telemedien", "impressum"),
("telekommunikationsgesetz", "impressum"),
("tkg", "impressum"),
("tmg", "impressum"),
("mediengesetz", "impressum"),
("gewerbeordnung", "impressum"),
("e-commerce", "agb"),
("digitale-inhalte", "agb"),
("konsumentenschutz", "verbraucherschutz"),
("urheberrecht", "urheberrecht"),
("urhg", "urheberrecht"),
("uwg", "wettbewerbsrecht"),
("handelsgesetzbuch", "handelsrecht"),
("hgb", "handelsrecht"),
("ugb", "handelsrecht"),
("abgb", "handelsrecht"),
("bgb", "agb"),
("gleichbehandlung", "gleichbehandlung"),
("(agg)", "gleichbehandlung"),
("abgabenordnung", "steuerrecht"),
("bao", "steuerrecht"),
("standardvertragsklauseln", "avv"),
("(scc)", "avv"),
# Datenschutz-Catch-alls (zuletzt)
("nist privacy framework", "dse"),
("dsgvo", "dse"),
("datenschutzgesetz", "dse"),
("bdsg", "dse"),
("edpb", "dse"),
("edps", "dse"),
("dsk ", "dse"),
("wp29", "dse"),
("bfdi", "dse"),
("data privacy framework", "dse"),
("datenschutz", "dse"),
)
def use_case_for_regulation(regulation: str | None) -> str | None:
"""Deterministisch: Quell-Regulierung → Domaenen-Use-Case (erster
Substring-Treffer). None wenn keine Regel passt (→ Fallback/Review)."""
if not regulation:
return None
low = regulation.lower()
for needle, uc in _REGULATION_RULES:
if needle in low:
return uc
return None
REGISTRY: dict[str, UseCase] = {uc.key: uc for uc in _USE_CASES}
@@ -0,0 +1,40 @@
-- Migration 150: Primaerzweck-Flag + MC <-> Quell-Regulierung (n:m)
-- Erweitert das Use-Case-Mapping um (a) is_primary (Primaerzweck pro MC)
-- und (b) mc_regulations (die feine Quell-Regulierungs-Filter-Dimension,
-- 117 Werte). Strikt add-only. [migration-approved]
SET search_path TO compliance, public;
DO $$
BEGIN
IF EXISTS (SELECT 1 FROM information_schema.tables
WHERE table_schema='compliance'
AND table_name='mc_use_case_mappings') THEN
ALTER TABLE mc_use_case_mappings
ADD COLUMN IF NOT EXISTS is_primary BOOLEAN NOT NULL DEFAULT FALSE;
CREATE INDEX IF NOT EXISTS idx_mcuc_primary
ON mc_use_case_mappings(use_case) WHERE is_primary;
END IF;
IF EXISTS (SELECT 1 FROM information_schema.tables
WHERE table_schema='compliance'
AND table_name='master_controls') THEN
CREATE TABLE IF NOT EXISTS mc_regulations (
id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
master_control_uuid UUID NOT NULL
REFERENCES master_controls(id) ON DELETE CASCADE,
master_control_id VARCHAR(60) NOT NULL,
source_regulation VARCHAR(160) NOT NULL,
is_primary BOOLEAN NOT NULL DEFAULT FALSE,
member_count INTEGER DEFAULT 0,
method VARCHAR(20) NOT NULL DEFAULT 'lineage'
CHECK (method IN ('lineage', 'manual')),
created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
UNIQUE (master_control_uuid, source_regulation)
);
CREATE INDEX IF NOT EXISTS idx_mcreg_regulation
ON mc_regulations(source_regulation);
CREATE INDEX IF NOT EXISTS idx_mcreg_mc
ON mc_regulations(master_control_uuid);
END IF;
END $$;
@@ -1,9 +1,12 @@
"""Klassifiziert Master Controls auf Use Cases (n:m) + Verifikations-Methode.
"""Klassifiziert Master Controls deterministisch auf Use Cases (n:m) +
Quell-Regulierung (n:m) + Verifikations-Methode.
Stufe 1 — Seed (kein LLM, gratis): aus vorhandenen Member-Signalen
(canonical_controls.scope_doc_type / .category / .verification_method /
.evidence_type) via `use_case_registry.seed_classify`.
Stufe 2 — LLM (Phase 3): Multi-Label gegen die Registry-Taxonomie. [TODO]
DETERMINISTISCH (kein LLM): Die Zuordnung kommt aus der Quell-Regulierung
jedes Controls — Lineage master_controls -> master_control_members ->
control_parent_links.source_regulation. 117 Regulierungen -> Keyword-Mapper
(use_case_registry.use_case_for_regulation) -> ~30 Domaenen-Use-Cases.
Primaerzweck = dominante Quell-Regulierung (meiste Member); Mehrfachzwecke =
die weiteren. LLM-Stufe (spaeter) nur Fallback fuer MCs ohne source_regulation.
Lauf im Container:
docker exec bp-compliance-backend \
@@ -15,83 +18,131 @@ from __future__ import annotations
import argparse
import asyncio
import os
import sys
import asyncpg
# /app auf den Pfad, damit `compliance...` als Standalone-Script importierbar
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from compliance.data import use_case_registry as reg
import asyncpg # noqa: E402
_AGG_SQL = """
from compliance.data import use_case_registry as reg # noqa: E402
_REG_SQL = """
SELECT mc.id AS mc_uuid, mc.master_control_id,
array_agg(DISTINCT cc.scope_doc_type) AS scopes,
array_agg(DISTINCT cc.category) AS categories,
array_agg(DISTINCT cc.verification_method) AS vmethods,
array_agg(DISTINCT cc.evidence_type) AS etypes
cpl.source_regulation AS reg, count(*) AS cnt
FROM compliance.master_controls mc
JOIN compliance.master_control_members mcm
ON mcm.master_control_uuid = mc.id
JOIN compliance.master_control_members mcm ON mcm.master_control_uuid = mc.id
JOIN compliance.control_parent_links cpl ON cpl.control_uuid = mcm.control_uuid
WHERE cpl.source_regulation IS NOT NULL AND cpl.source_regulation <> ''
GROUP BY mc.id, mc.master_control_id, cpl.source_regulation
"""
_VERIF_SQL = """
SELECT mc.id AS mc_uuid,
array_agg(DISTINCT cc.verification_method) AS vmethods,
array_agg(DISTINCT cc.evidence_type) AS etypes
FROM compliance.master_controls mc
JOIN compliance.master_control_members mcm ON mcm.master_control_uuid = mc.id
JOIN compliance.canonical_controls cc ON cc.id = mcm.control_uuid
GROUP BY mc.id, mc.master_control_id
GROUP BY mc.id
"""
async def run_seed(conn, limit: int = 0) -> dict:
"""Deterministischer Seed → mc_use_case_mappings + mc_verification.
Idempotent (ON CONFLICT DO NOTHING); ueberschreibt 'manual' nie."""
sql = _AGG_SQL + (f" LIMIT {limit}" if limit > 0 else "")
rows = await conn.fetch(sql)
n_mc_with_uc = n_uc_rows = n_verif = 0
for r in rows:
ucs, method = reg.seed_classify(
r["scopes"], r["categories"], r["vmethods"], r["etypes"],
)
for uc in ucs:
"""Deterministischer Seed aus der Quell-Regulierung. Ersetzt die
bisherigen Seed-Zeilen; 'manual'-Korrekturen bleiben unangetastet."""
await conn.execute(
"DELETE FROM compliance.mc_use_case_mappings WHERE method='seed'")
await conn.execute(
"DELETE FROM compliance.mc_verification WHERE method='seed'")
await conn.execute(
"DELETE FROM compliance.mc_regulations WHERE method='lineage'")
by_mc: dict = {}
for r in await conn.fetch(_REG_SQL):
e = by_mc.setdefault(
r["mc_uuid"],
{"mc_id": r["master_control_id"], "regs": {}})
e["regs"][r["reg"]] = r["cnt"]
if limit > 0:
by_mc = dict(list(by_mc.items())[:limit])
verif: dict = {}
for r in await conn.fetch(_VERIF_SQL):
_ucs, m = reg.seed_classify(vmethods=r["vmethods"], etypes=r["etypes"])
if m:
verif[r["mc_uuid"]] = m
n_reg = n_uc = n_mc = n_v = 0
for mc_uuid, e in by_mc.items():
mc_id, regs = e["mc_id"], e["regs"]
if not regs:
continue
primary_reg = max(regs, key=regs.get)
uc_primary = reg.use_case_for_regulation(primary_reg)
ucs: dict = {}
for rg, cnt in regs.items():
await conn.execute(
"""INSERT INTO compliance.mc_regulations
(master_control_uuid, master_control_id, source_regulation,
is_primary, member_count)
VALUES ($1,$2,$3,$4,$5)
ON CONFLICT (master_control_uuid, source_regulation)
DO NOTHING""",
mc_uuid, mc_id, rg[:160], rg == primary_reg, cnt)
n_reg += 1
uc = reg.use_case_for_regulation(rg)
if uc:
ucs[uc] = ucs.get(uc, False) or (uc == uc_primary)
for uc, is_prim in ucs.items():
await conn.execute(
"""INSERT INTO compliance.mc_use_case_mappings
(master_control_uuid, master_control_id, use_case,
method, confidence, rationale)
VALUES ($1,$2,$3,'seed',0.6,'deterministic seed')
ON CONFLICT (master_control_uuid, use_case) DO NOTHING""",
r["mc_uuid"], r["master_control_id"], uc,
)
n_uc_rows += 1
if ucs:
n_mc_with_uc += 1
if method:
method, confidence, rationale, is_primary)
VALUES ($1,$2,$3,'seed',0.85,'source_regulation',$4)
ON CONFLICT (master_control_uuid, use_case)
DO UPDATE SET is_primary =
mc_use_case_mappings.is_primary OR EXCLUDED.is_primary
WHERE mc_use_case_mappings.method <> 'manual'""",
mc_uuid, mc_id, uc, is_prim)
n_uc += 1
n_mc += 1
m = verif.get(mc_uuid)
if m:
await conn.execute(
"""INSERT INTO compliance.mc_verification
(master_control_uuid, master_control_id,
verification_method, method, confidence, rationale)
VALUES ($1,$2,$3,'seed',0.6,'deterministic seed')
VALUES ($1,$2,$3,'seed',0.7,'member evidence_type')
ON CONFLICT (master_control_uuid) DO NOTHING""",
r["mc_uuid"], r["master_control_id"], method,
)
n_verif += 1
mc_uuid, mc_id, m)
n_v += 1
total = await conn.fetchval(
"SELECT count(*) FROM compliance.mc_use_case_mappings")
await conn.execute(
"""INSERT INTO compliance.mc_use_case_sync_state
(registry_hash, stage, total_mappings, mcs_classified)
VALUES ($1,'seed',$2,$3)""",
reg.registry_hash(), total, n_mc_with_uc,
)
return {"mcs": len(rows), "mcs_with_use_case": n_mc_with_uc,
"use_case_rows": n_uc_rows, "verification_rows": n_verif}
VALUES ($1,'seed_source_regulation',$2,$3)""",
reg.registry_hash(), total, n_mc)
return {"mcs_mapped": n_mc, "regulation_rows": n_reg,
"use_case_rows": n_uc, "verification_rows": n_v}
async def _main() -> None:
ap = argparse.ArgumentParser()
ap.add_argument("--limit", type=int, default=0)
ap.add_argument("--with-llm", action="store_true",
help="Phase 3 — noch nicht implementiert")
help="Fallback fuer MCs ohne source_regulation (TODO)")
args = ap.parse_args()
if args.with_llm:
raise SystemExit("LLM-Stufe (Phase 3) noch nicht implementiert.")
raise SystemExit("LLM-Fallback (Phase 3) noch nicht implementiert.")
conn = await asyncpg.connect(os.getenv("DATABASE_URL"))
try:
stats = await run_seed(conn, args.limit)
finally:
await conn.close()
print("Seed fertig:", stats)
print("Seed (source_regulation) fertig:", stats)
if __name__ == "__main__":
@@ -110,3 +110,38 @@ def test_seed_empty_and_none_safe():
vmethods=[None], etypes=[None])
assert ucs == [] and m is None
assert reg.seed_classify() == ([], None)
# ── Deterministischer Regulierung→Use-Case-Mapper ───────────────────
def test_regulation_mapper_known():
cases = {
"OWASP Top 10 (2021)": "code_security",
"NIST SP 800-207 (Zero Trust)": "network_security",
"Cyber Resilience Act (CRA)": "cra",
"DSGVO (EU) 2016/679": "dse",
"EDPB Facial Recognition": "dse", # Leitlinie → Datenschutz
"TKG": "impressum",
"TDDDG": "cookie_banner",
"Markets in Crypto-Assets (MiCA)": "mica",
"BGB": "agb",
}
for reg_str, expected in cases.items():
assert reg.use_case_for_regulation(reg_str) == expected, reg_str
def test_regulation_mapper_abgb_before_bgb():
# 'ABGB' enthaelt 'bgb' — die abgb-Regel MUSS zuerst greifen.
assert reg.use_case_for_regulation("AT ABGB") == "handelsrecht"
def test_regulation_mapper_unknown_returns_none():
assert reg.use_case_for_regulation("Voellig Unbekanntes Gesetz") is None
assert reg.use_case_for_regulation(None) is None
def test_all_regulation_rules_point_to_valid_use_cases():
for _needle, uc in reg._REGULATION_RULES:
assert uc in reg.REGISTRY, uc
assert reg.REGISTRY[uc].enabled