cosmix/loom

loom-threat-model

Threat modeling methodologies (STRIDE, DREAD, PASTA, attack trees) for secure architecture design.

소스 보기

사용 방법

  1. 설치 명령을 복사하고 자신의 에이전트 환경에 설치하기 전에 소스 패키지를 확인하세요.
  2. 필요한 도구, 인증 정보, 지원 환경은 원본 문서에서 확인하세요.
  3. 입력과 원하는 산출물을 설명하고 출처와 결과를 대조하세요.
npx skills add https://github.com/cosmix/loom --skill loom-threat-model

시작 전 확인

  • 설치에는 모델 구독, API 한도, 파일 접근 권한이 포함되지 않습니다.
  • 별점 수와 설치 수는 활동 지표이며 독립적인 품질·보안 검증이 아닙니다.
원본 Skill 문서

원본 저장소의 제목, 예시, 코드, 표, 링크, 이미지를 유지해 표시합니다.

Threat Modeling

Structured identification of what can go wrong in a design, before code exists. Answers four questions (Shostack): What are we building? What can go wrong? What are we doing about it? Did we do a good job? This skill is architecture-time analysis — for finding vulns in existing code use loom-security-scan/loom-security-audit; for auth mechanism details use loom-auth.

When

New system design, architecture review, significant feature or trust-boundary change, third-party integration, or compliance evidence. Re-run when the architecture changes — a threat model is a living document, not a one-time deliverable.

Methodologies

STRIDE — the default; apply per element of the DFD

The core technique isn't "brainstorm STRIDE" — it's walking each DFD element and each data flow crossing a trust boundary, asking which STRIDE categories apply to that element.

ThreatViolatesTypical control
SpoofingAuthenticationStrong authn, mTLS, signed tokens
TamperingIntegritySignatures, HMAC, input validation, WORM logs
RepudiationNon-repudiationAudit logs, signed receipts
Information disclosureConfidentialityEncryption, least-privilege, error hygiene
Denial of serviceAvailabilityRate limits, quotas, timeouts, autoscale
Elevation of privilegeAuthorizationAuthZ checks, sandboxing, least privilege

Element→likely-STRIDE heuristic: external entities → S, R; processes → all six; data flows → T, I, D; data stores → T, I, D (and R if logs).

DREAD — risk scoring (use with caution)

Score Damage, Reproducibility, Exploitability, Affected users, Discoverability (1–10); risk = mean. ⚠ DREAD is widely criticized as subjective and inconsistent across raters (Microsoft dropped it). Prefer a simple Likelihood × Impact matrix, or CVSS for concrete vulns, when you need defensible numbers. Whatever the scale, rank threats to drive mitigation order.

PASTA / LINDDUN

  • PASTA — 7-stage, risk-centric, business-objective-driven (objectives → tech scope → decomposition → threat analysis → vuln analysis → attack modeling → risk & mitigation). Use for high-stakes systems needing business alignment.
  • LINDDUN — the STRIDE-equivalent for privacy threats (Linkability, Identifiability, Non-repudiation, Detectability, Disclosure, Unawareness, Non-compliance). Reach for it on GDPR/PII-heavy systems.

Process

1. Scope & assets

List assets and why an attacker wants them — this drives everything.

AssetClassificationImpact if compromised
Credentials/secretsConfidentialAccount/system takeover
Payment dataPCI-DSSFinancial + compliance loss
PIIGDPRPrivacy breach, fines

2. DFD with trust boundaries (the load-bearing step)

Threats concentrate at trust boundaries — every arrow crossing one is where authn/authz/validation must live. Mark: external entities, processes, data stores, data flows, and boundaries (internet↔DMZ, DMZ↔internal, tenant↔tenant, host↔container, user↔kernel).

text
[User] --HTTPS--> ║ DMZ ║ [API Gateway] --> ║ Internal ║ [App] --> [DB]
                  ↑ boundary: authn, TLS       ↑ boundary: authz, network policy, mTLS

A boundary you didn't draw is a check you won't add. Enumerate every flow: source, destination, protocol, data classification, boundary crossed.

3. Enumerate threats (STRIDE per element)

markdown
### API Gateway
| STRIDE | Threat | Likelihood | Impact |
| ------ | ------ | ---------- | ------ |
| S | Forged/`alg:none` JWT | Med | High |
| I | Verbose error leaks stack/version | High | Med |
| D | Rate-limit bypass | Med | High |
| E | BOLA/IDOR → other users' objects | High | Critical |

4. Attack trees (for high-value goals)

Decompose a goal into OR/AND paths; the cheapest leaf is the likely attack and shows where to spend defense.

text
             Steal User Data (goal)
        ┌──────────┼──────────┐        (OR)
   Compromise    Exploit App   Social
   Credentials   Vulnerability Engineering
    ┌───┴───┐     ┌───┴───┐
 Phishing Brute  SQLi   BOLA
[L:H I:H][L:M]  [L:M I:C][L:H I:H]

5. Mitigations — track to closure

markdown
| Threat | Mitigation | Priority | Status |
| ------ | ---------- | -------- | ------ |
| BOLA/IDOR | Object-level ownership checks (loom-auth) | P0 | Not started |
| SQLi | Parameterized queries | P0 | In progress |
| JWT forgery | Pin alg, validate aud/iss/exp (loom-auth) | P1 | Done |

Every high/critical threat needs a mitigation, an accepted-risk decision (with owner + justification), or a transfer. Silence = unmanaged risk.

High-Signal Threats & Their Real Controls

  • BOLA / IDOR (broken object-level authz) — #1 API threat: any object accessed by id must verify caller ownership/tenant, not just "logged in". Details + wrong-vs-right in loom-auth.
  • SSRF — attacker makes your server fetch an internal URL (cloud metadata 169.254.169.254, localhost, internal services). Defense is an allowlist of permitted hosts/schemes — never a blocklist. Blocklists are bypassed by DNS-rebinding, HTTP redirects, IPv6 ([::1]), and decimal/octal/hex-encoded IPs (http://2130706433/). Also disable unused URL schemes and block redirects to non-allowlisted hosts.
  • Injection (SQL/NoSQL/command/LDAP/XXE) — parameterize/allowlist; treat every trust-boundary input as hostile.
  • Mass assignment — client sets fields it shouldn't (role, isAdmin, tenant_id); bind explicit allowlists.
  • Cloud metadata & IAM escalation — SSRF-to-credentials via the metadata endpoint; enforce IMDSv2, least-privilege roles.
  • Supply chain — dependency confusion, typosquat, unsigned artifacts (see loom-dependency-scan).

Component Quick-Reference

ComponentWatch for
Web appXSS, CSRF, clickjacking, open redirect, session fixation
APIBroken authn/authz (BOLA/BFLA), mass assignment, rate-limit bypass, injection
DatabaseSQLi, privilege escalation, unencrypted data/backups
AuthCredential stuffing, session fixation, token leakage, MFA bypass
File uploadMalware, path traversal, RCE via content, storage exhaustion
Cloud/infraPublic buckets, IAM escalation, SSRF→metadata, exposed control plane
Containers/K8sVulnerable base images, secrets in env/layers, privileged pods, RBAC gaps, container escape
3rd-party/webhooksAPI-key exposure, webhook spoofing (verify HMAC), supply-chain
ML systemsData poisoning, evasion/adversarial input, model inversion/extraction, PII leakage in outputs

Threat Model Document Template

markdown
# Threat Model: <System>   (v1.0, YYYY-MM-DD, author, reviewers)
1. Executive summary — key risks & recommendations
2. System description — purpose, architecture, data classification, trust boundaries
3. Assets — table (asset, classification, owner)
4. Analysis — DFD, STRIDE-per-component, attack trees
5. Risk assessment — table (threat, likelihood×impact or CVSS, level)
6. Mitigations — table (threat, mitigation, owner, timeline, status)
7. Residual risks — accepted risks + justification
8. Review schedule — trigger conditions to revisit

Verification Checklist

  • [ ] DFD drawn with every trust boundary marked; each boundary-crossing flow enumerated
  • [ ] STRIDE applied per element, not brainstormed globally
  • [ ] Assets tied to attacker motivation and classification
  • [ ] Every high/critical threat has a mitigation, accepted-risk, or transfer with an owner
  • [ ] BOLA/IDOR checked on every object-addressing endpoint; SSRF defenses are allowlist-based
  • [ ] Risk ranking uses a defensible scale (Likelihood×Impact or CVSS, not raw DREAD gut-feel)
  • [ ] Model versioned and has a re-review trigger; privacy threats covered (LINDDUN) if PII-heavy
같은 저장소의 Skills

더 많은 Skills

모든 Skills