Skip to content

Risk Register

IEC 62443-4-1 SM-5 Compliance

1. Purpose

This register tracks all identified security and operational risks for the RUNE platform. It is the primary output of the risk assessment methodology defined in RISK_ASSESSMENT.md. Each risk is scored, assigned a treatment, and reviewed on a regular cadence.

2. Risk Register

ID Category Description L I RS Treatment Owner Status Review Date
R-001 Operational / Cost Vast.ai GPU provisioning costs exceed budget due to misconfigured or runaway instances. Fail-closed cost gate mitigates but does not eliminate risk for manual operations. 3 4 12 Mitigate lpasquali Open 2026-07-01
R-002 Supply Chain Dependency confusion or typosquatting attack introduces malicious package into requirements.txt or go.mod. 2 5 10 Mitigate lpasquali Open 2026-07-01
R-003 Supply Chain CI pipeline compromise (GitHub Actions workflow injection, compromised action). 2 5 10 Mitigate lpasquali Open 2026-07-01
R-004 Information Disclosure Secrets (API tokens, Vast.ai keys) leaked via logs, error messages, or misconfigured environment. Policy mandates env-var-only storage, but human error remains possible. 3 4 12 Mitigate lpasquali Open 2026-07-01
R-005 Tampering Unsigned container images accepted by Kubernetes cluster, allowing deployment of tampered artifacts. See IMAGE_SIGNING.md for planned mitigation. 3 5 15 Mitigate lpasquali Open 2026-07-01
R-006 Elevation of Privilege Kubernetes RBAC misconfiguration in rune-operator grants excessive permissions to workloads. 2 4 8 Mitigate lpasquali Open 2026-07-01
R-007 Denial of Service DriverTransport protocol lacks rate limiting; malicious or buggy driver could flood the orchestrator. 3 3 9 Mitigate lpasquali Open 2026-07-01
R-008 Compliance / Legal GPL-2.0 licensed dependency introduced without CI detection. License gate updated to block GPL-2.0 variants (see quality-gates.yml). 2 5 10 Mitigate lpasquali Mitigated 2026-07-01
R-009 Information Disclosure SQLite database in Kubernetes lacks encryption at rest; persistent volume could be accessed if node is compromised. 2 3 6 Accept lpasquali Accepted 2026-07-01
R-010 Spoofing Unsigned Git commits could be injected if branch protection is misconfigured. Signed commits are required by policy but not enforced by all repos yet. 2 4 8 Mitigate lpasquali Open 2026-07-01
R-011 Tampering LLM response injection -- malicious LLM output could manipulate agent behavior or produce misleading benchmark results. 3 3 9 Mitigate lpasquali Open 2026-07-01
R-012 Repudiation Incomplete audit trail for Vast.ai instance lifecycle events; manual operations are not logged. 3 3 9 Mitigate lpasquali Open 2026-07-01
R-013 Denial of Service MkDocs documentation site denial of service via resource exhaustion (low impact; static site behind CDN). 1 1 1 Accept lpasquali Accepted 2026-07-01
R-014 Supply Chain Container base image (Python/Go) contains unpatched CVEs. Mitigated by SBOM + Grype/Trivy scanning in CI; unfixable CVEs tracked in VEX. 3 3 9 Mitigate lpasquali Open 2026-07-01
R-015 Compliance IEC 62443-4-1 certification evidence gaps. This risk register and supporting security documents are being developed to address the documentation gaps. 2 4 8 Mitigate lpasquali In Progress 2026-07-01

3. Risk Score Summary

quadrantChart
    title Risk Heatmap
    x-axis Low Impact --> High Impact
    y-axis Low Likelihood --> High Likelihood
    quadrant-1 Monitor Closely
    quadrant-2 Immediate Action
    quadrant-3 Accept & Review
    quadrant-4 Plan Mitigation
    R-001 Vast.ai Costs: [0.75, 0.55]
    R-002 Dep Confusion: [0.95, 0.35]
    R-003 CI Compromise: [0.95, 0.35]
    R-004 Secret Leakage: [0.75, 0.55]
    R-005 Unsigned Images: [0.95, 0.55]
    R-007 DT Rate Limit: [0.55, 0.55]
    R-008 GPL License: [0.95, 0.35]
    R-011 LLM Injection: [0.55, 0.55]

4. Treatment Plan Summary

Treatment Count Risk IDs
Mitigate 12 R-001 through R-008, R-010, R-011, R-012, R-014
Accept 2 R-009, R-013
In Progress 1 R-015

5. Review Schedule

  • Quarterly review: All open risks reassessed for likelihood and impact changes.
  • Post-incident: Any risk involved in a P0/P1 incident is immediately reassessed.
  • New risk intake: Risks identified during penetration tests (PENTEST.md), fuzz testing (FUZZ_TESTING.md), or threat modeling are added within 5 business days.

6. References