🛡️ SHACL Validation Profiles

Below are production-grade SHACL validation profiles that enforce ontology purity, extension boundaries, and RAG-safe traversals.

These shapes are not descriptive — they are guardrails. If data passes SHACL, it is safe to reason over.

Overview (Cybersecurity Ontology Governance Layer)

Applies to:

🔐 Core Ontology v1.0
🚨 Incident Extension
⚖️ Risk Extension
🎭 ThreatActor Extension

Design goal:

❌ Block invalid facts ❌ Block semantic pollution ❌ Block hallucination-enabling shortcuts

0. SHACL Strategy (How to Use This)

You should run SHACL validation at three points:

Ingest time (ETL / pipeline)
Pre-RAG index build
Pre-query execution (optional but powerful)

If a shape fails → ➡️ reject data or downgrade confidence

1. Core Ontology SHACL Profiles

SH-CORE-01 — Vulnerability Must Have Weakness

Enforces the causal backbone No CVE without a root cause

@prefix sh: <http://www.w3.org/ns/shacl#> .
@prefix sec: <http://example.org/sec#> .

sec:VulnerabilityShape
  a sh:NodeShape ;
  sh:targetClass sec:Vulnerability ;
  sh:property [
    sh:path sec:caused_by ;
    sh:minCount 1 ;
    sh:message "Every Vulnerability must be linked to at least one Weakness (CWE)." ;
  ] .

SH-CORE-02 — No CVE → Technique Shortcut

sec:NoDirectVulnTechniqueShape
  a sh:NodeShape ;
  sh:targetClass sec:Vulnerability ;
  sh:not [
    sh:property [
      sh:path sec:implemented_as ;
      sh:minCount 1 ;
    ]
  ] ;
  sh:message "Vulnerability must not directly link to Technique." .

SH-CORE-03 — Technique Must Belong to a Tactic

sec:TechniqueTacticShape
  a sh:NodeShape ;
  sh:targetClass sec:Technique ;
  sh:property [
    sh:path sec:belongs_to ;
    sh:minCount 1 ;
    sh:message "Every Technique must belong to at least one Tactic." ;
  ] .

2. Incident Extension SHACL Profiles

SH-INC-01 — Incident Must Have Observations

@prefix inc: <http://example.org/incident#> .

inc:IncidentObservationShape
  a sh:NodeShape ;
  sh:targetClass inc:Incident ;
  sh:property [
    sh:path inc:observed ;
    sh:minCount 1 ;
    sh:message "Incident must include at least one ObservedTechnique." ;
  ] .

SH-INC-02 — ObservedTechnique Must Anchor to Core

Prevents fabricated behavior

inc:ObservedTechniqueAnchorShape
  a sh:NodeShape ;
  sh:targetClass inc:ObservedTechnique ;
  sh:property [
    sh:path inc:instance_of ;
    sh:class sec:Technique ;
    sh:minCount 1 ;
    sh:maxCount 1 ;
    sh:message "ObservedTechnique must reference exactly one Core Technique." ;
  ] .

SH-INC-03 — No Incident → Vulnerability Link

inc:NoIncidentVulnerabilityShape
  a sh:NodeShape ;
  sh:targetClass inc:Incident ;
  sh:not [
    sh:property [
      sh:path sec:affected_by ;
      sh:minCount 1 ;
    ]
  ] ;
  sh:message "Incident must not directly reference Vulnerability." .

3. ThreatActor Extension SHACL Profiles

(Attribution Without Pollution)

SH-TA-01 — Attribution Must Be Claim-Based

@prefix ta: <http://example.org/threatactor#> .

ta:AttributionClaimShape
  a sh:NodeShape ;
  sh:targetClass ta:AttributionClaim ;
  sh:property [
    sh:path ta:attributes ;
    sh:minCount 1 ;
    sh:message "AttributionClaim must reference a ThreatActor." ;
  ] ;
  sh:property [
    sh:path ta:based_on ;
    sh:minCount 1 ;
    sh:message "AttributionClaim must be based on an Incident." ;
  ] ;
  sh:property [
    sh:path ta:confidence ;
    sh:minCount 1 ;
    sh:message "AttributionClaim must include confidence." ;
  ] .

SH-TA-02 — No Direct Incident → ThreatActor

ta:NoDirectIncidentActorShape
  a sh:NodeShape ;
  sh:targetClass inc:Incident ;
  sh:not [
    sh:property [
      sh:path ta:attributes ;
      sh:minCount 1 ;
    ]
  ] ;
  sh:message "Incident must not directly reference ThreatActor." .

SH-TA-03 — ThreatActors Operate Only via Techniques

ta:ThreatActorTechniqueOnlyShape
  a sh:NodeShape ;
  sh:targetClass ta:ThreatActor ;
  sh:not [
    sh:property [
      sh:path sec:caused_by ;
      sh:minCount 1 ;
    ]
  ] ;
  sh:message "ThreatActor must not reference Weakness or Vulnerability directly." .

4. Risk Extension SHACL Profiles

(Decision Discipline)

SH-RISK-01 — RiskScenario Must Anchor to Core

@prefix risk: <http://example.org/risk#> .

risk:RiskScenarioAnchorShape
  a sh:NodeShape ;
  sh:targetClass risk:RiskScenario ;
  sh:property [
    sh:path risk:considers ;
    sh:minCount 1 ;
    sh:message "RiskScenario must consider at least one Core entity." ;
  ] .

SH-RISK-02 — Likelihood Requires Rationale

risk:LikelihoodRationaleShape
  a sh:NodeShape ;
  sh:targetClass risk:Likelihood ;
  sh:property [
    sh:path risk:rationale ;
    sh:minCount 1 ;
    sh:datatype xsd:string ;
    sh:message "Likelihood must include rationale." ;
  ] .

SH-RISK-03 — Risk Must Not Override Core Severity

risk:NoCVSSOverrideShape
  a sh:NodeShape ;
  sh:targetClass risk:RiskScore ;
  sh:not [
    sh:property [
      sh:path sec:scored_by ;
      sh:minCount 1 ;
    ]
  ] ;
  sh:message "RiskScore must not modify CVSS or Core severity." .

5. RAG-Specific Guardrail Shapes

SH-RAG-01 — No Forbidden Shortcuts

sec:NoForbiddenEdgesShape
  a sh:NodeShape ;
  sh:targetSubjectsOf sec:affected_by ;
  sh:not [
    sh:class ta:ThreatActor ;
  ] ;
  sh:message "Forbidden traversal detected: Vulnerability → ThreatActor." .

SH-RAG-02 — Observations Must Be Evidence-Backed

inc:ObservationEvidenceShape
  a sh:NodeShape ;
  sh:targetClass inc:ObservedTechnique ;
  sh:property [
    sh:path inc:evidenced_by ;
    sh:minCount 1 ;
    sh:message "ObservedTechnique must be supported by Evidence." ;
  ] .

6. Enforcement Levels (Recommended)

Level	When	Action
STRICT	Ingest	Reject data
STANDARD	Pre-RAG	Exclude nodes
LENIENT	Exploration	Flag with warnings

7. What These SHACL Profiles Guarantee

✅ No hallucinated causality ✅ No attribution leakage ✅ No risk subjectivity in Core ✅ No unsafe RAG traversal ✅ Audit-ready reasoning