User-Agent Remote Command Injection

This project demonstrates how an insecure use of the User-Agent HTTP header can lead to Remote Command Execution (RCE) if not properly sanitised. It includes:

• A vulnerable PHP server that logs headers unsafely using system()
• A secure version with input sanitation
• A Python-based exploit tool that simulates an attack
• Defensive guidelines and mitigation recommendations

Author

GitHub: BackdoorAli

Exploit Summary

When a web server logs the User-Agent header directly into a system shell command, it becomes vulnerable to injection if the attacker includes malicious shell syntax in the header.

Example malicious header:

User-Agent: zerodium; id

In an insecure environment, this results in execution of the id command on the server.

Project Files

• vulnerable_index.php — Insecure PHP script (for demo only)
• secure_index.php — Hardened PHP script using escapeshellarg()
• exploit.py — CLI tool to send injection payloads via User-Agent header

Usage

Start the Vulnerable Server

cd vulnerable_server
php -S 127.0.0.1:8000

Run the Exploit

python3 exploit.py http://127.0.0.1:8000/ "id"

Output (vulnerable server):

[+] Response Body:
uid=501(alita) gid=20(staff) groups=...

Try the Secure Server

cd secure_server
php -S 127.0.0.1:8000

Re-running the same exploit will now produce no harmful effects.

---

Disclaimer

This project is for educational and awareness purposes only. Do NOT deploy the vulnerable server in a production environment or expose it to the internet. Always sanitise user input and NEVER directly inject user data into system commands, without the proper authorisation!