Dangerous use of express body-parser

December 18, 2015 — Chris Foster

Cross-Site Request Forgery, or CSRF, is a type of attack that developers are familiar with in traditional web applications, but often misunderstand or forget about when it comes to new REST API’s. Fortunately, much of this misunderstanding and lack of consideration occurs because full page applications often don’t need to worry about CSRF. While many architectural differences in REST reduce the risk of CSRF attacks, that doesn’t mean we can simply ignore them entirely. Express’s body-parser module is a great recent example of this.

When you don’t need CSRF protection

With Javascript XHR requests, it’s not possible to make cross-origin requests the same way it is with forms due to the same-origin policy. If you attempt to make an AJAX request to another origin, the browser will disallow the request. HTML forms are not capable of generating anything other than form data, so they can’t talk to REST backends. Therefore since only XHR’s can send JSON data to an API, an API that accepts only JSON is protected from most CSRF attacks.

CSRF Diagram

When you do need CSRF protection

So how does all of this relate to express’s body-parser module? If you’re using Nodejs and express, there is a very high chance that your building a server side API. The de facto way to parse JSON in express is body-parser, and the code they always recommend using is:

var bodyParser = require('body-parser');
    extended: true

The problem with the above code is that if you use it, you need CSRF protection. In your route middleware, body-parser will map JSON data to the req.body object present in each request. However, when you call both bodyParser.json() and bodyParser.urlencoded() you are populating req.body with content from both JSON sources and form-encoded sources. Your API route handlers cannot tell the difference between what Content-Type populated the API.

This means that your API route could be called from either a Javascript XHR request or a traditional HTML form. If your application uses cookies, then that means that an attacker could create a malicious HTML form with form-encoded data that submits to your API, triggering whatever operation they’d like with the permissions of the currently logged in user. This is your traditional CSRF attack, but made possible because you parse more data types than you realize.

Making it safe

If your app has this, there are multiple ways you can fix this:

  • Use traditional CSRF tokens for all requests
  • Remove the bodyParser.urlencoded middleware from API routes
  • Check the Content-Type header of the request

Personally, I’d recommend removing the bodyParser.urlencoded middleware, since it’s the most straightforward in my opinion. If you parse any sort of traditional HTML form, such as for user login, then you will need to be add the bodyParser.urlencoded middleware support on that route, but keep it off all the others! The documentation covers how to do this.

Beware example code

In summary, this isn’t an actual vulnerability or something that is wrong with body-parser. The author provided defaults that would cover the use case for 99% of the middleware’s users. This is a configuration-based insecurity, and is great example of how even simple example code can lead to a vulnerability in your own application. Don’t trust code blindly, be sure to understand what you’re putting in your application and how it affects your security.

This is a side effect of my recent messing around with CSRF. If you’d like to see more articles like this, follow me on Twitter.

Back to the index page