Overview
A decades-frozen asset has been reintegrated, and the records system tracking the process is exposed as a thin Flask service behind nginx. The landing page advertises a GraphQL endpoint and three role-scoped namespaces: /field/, /division/ and /director/.
The chain is deliberately linear — each stage hands over the privilege needed by the next:
GraphQL introspection
-> field token -> FLAG 1
IDOR on /field/dossier/<id> -> FLAG 2
requestAuditToken + aud-skip -> division token -> FLAG 3
yaml.UnsafeLoader on /division/import -> RCE -> FLAG 4
SSRF into quartermaster-internal -> HMAC key -> FLAG 5
TOCTOU race on /approve -> director token -> FLAG 6
Prerequisites / Setup
export TARGET="<challenge URL or IP>"
curl is sufficient; no wordlist or credentials are required.
Recon
Step 1 — Inspect the application
Goal: Confirm the GraphQL endpoint and role-scoped namespaces.
curl -s "$TARGET/"
{
"api": "/graphql",
"namespaces": ["/field/", "/division/", "/director/"],
"service": "The Centurion Archive"
}
Every namespace without a token leaks its requirement but nothing else.
Vulnerability Analysis
The flaw classes are GraphQL Introspection (information disclosure), IDOR (Improper Access Control), YAML Deserialization (CWE-502), SSRF (Server-Side Request Forgery), and TOCTOU Race Condition.
Exploitation
Step 1 — GraphQL introspection
Goal: Discover the GraphQL schema and available mutations.
curl -s -X POST "$TARGET/graphql" \
-H "Content-Type: application/json" \
-d '{"query": "{ __schema { types { name kind fields { name type { name kind } } } } }"}'
Introspection returns a hand-written schema listing three mutations: register, requestFieldAccess, and requestAuditToken. Note that requestAuditToken takes a department argument rather than a password.
Step 2 — Field access token (Stage 1)
Goal: Obtain the field token via GraphQL introspection and IDOR.
Query the GraphQL schema for field definitions, then use IDOR on /field/dossier/<id> to access another user’s dossier and obtain the field token.
Flag
r00t{REDACTED}
Lessons / Takeaways
- Never expose GraphQL introspection in production without authentication.
- IDOR on object identifiers allows accessing another user’s resources.
yaml.UnsafeLoaderallows arbitrary Python object construction and code execution.- SSRF into internal services can leak sensitive keys and credentials.
- TOCTOU races between authorization check and resource use can be exploited for privilege escalation.
Mitigations
- Disable GraphQL introspection in production or restrict it to authenticated administrators.
- Implement proper authorization checks on all object access (IDOR prevention).
- Use
yaml.safe_load()instead ofyaml.UnsafeLoader. - Implement SSRF protection by validating and restricting outbound request targets.
- Use atomic operations or file locking to prevent TOCTOU races.
Techniques
- GraphQL Introspection
- IDOR
- YAML Deserialization
- SSRF
- TOCTOU
References
- OWASP, “Broken Access Control”: https://owasp.org/Top10/A01_2021-Broken_Access_Control/
- OWASP, “Server-Side Request Forgery”: https://owasp.org/www-community/attacks/Server_Side_Request_Forgery
- CWE-22: Path Traversal: https://cwe.mitre.org/data/definitions/22.html
- CWE-502: Deserialization of Untrusted Data: https://cwe.mitre.org/data/definitions/502.html

