|Join hundreds of InfoSec professionals at our upcoming |
[Global AppSec Amsterdam, September 23-27]
Anti CSRF Tokens ASP.NET
In short, CSRF abuses the trust relationship between browser and server. This means that anything that a server uses in order to establish trust with a browser (e.g., cookies, but also HTTP/Windows Authentication) is exactly what allows CSRF to take place - but this only the first piece for a successful CSRF attack.
The second piece is a web form or request which contains parameters predictable enough that an attacker could craft his own malicious form/request which, in turn, would be successfully accepted by the target service. Then, usually through social engineering or XSS, the victim would trigger that malicious form/request submission while authenticated to the legitimate service. This is where the browser/server trust is exploited.
In order to prevent CSRF in ASP.NET, anti-forgery tokens (also known as request verification tokens) must be utilized.
These tokens are simply randomly-generated values included in any form/request that warrants protection. Note that this value should be unique for every individual session. This guarantees that every form/request is tied to the authenticated user and, therefore, protected from CSRF.
Important: non-idempotent GET requests represent an anti-pattern where CSRF protection is concerned. Always use POST requests with anti-CSRF tokens for proper protection.
Please note that the following examples may not entail a complete anti-CSRF solution for any given Web application. Specific requirements may call for adjustments and/or combinations of different strategies.
Solutions NOT considered secure
- All of the solutions provided in this article are not designed to work with GET requests that change the server state (e.g., /example/delete.aspx?id=1). GET requests should be idempotent so that CSRF cannot take place.
- Assuming that SSL/TLS will thwart CSRF attacks just because the cookie is marked "Secure" and/or "HTTPOnly". The problem lies in the trust between legitimate browser and server. Therefore, the browser will just send its current cookies when the forged request is triggered. The attacker never has to touch any cookies.
- Referer header verification as the only protection. This can be easily manipulated.
- Any CSRF protection is null and void given the presence of XSS, for several reasons. The main and obvious reason is that, through XSS, the attacker can hijack the session and spoof the user, not even having to worry about performing CSRF.
ASP.NET MVC and Web API: Anti-CSRF Token
ASP.NET has the capability to generate anti-CSRF security tokens for consumption by your application, as such:
1) Authenticated user (has session which is managed by the framework) requests a page which contains form(s) that changes the server state (e.g., user options, account transfer, file upload, admin functions, etc.)
2) Generate the security token (or grab it from the session state) and send the token as a session cookie (again, managed in the session state, unique per session) as well as within a hidden value in each form.
3) Once the user submits the form, validate the token stored in the session state against the token included in the submitted form value. On failure, disregard form.
Effectively, this is the cookie double-submission approach done right, since the security token is submitted both as a cookie (managed in the framework session state) and within a hidden form value at the same time.
The standard frequency of token generation is per-session, so make sure your sessions have a reasonable/configurable time-out.
It is possible to issue new tokens on a per-request basis. However, the added protection may be insignificant, if this approach even fits your application. See the link below for a discussion on the matter:
Why refresh CSRF token per form request?
EnableViewStateMac must be set.
In fact, the ViewState MAC can no longer be disabled for versions since September 2014.
The ASP.NET ViewState contains a property, ViewStateUserKey, which offers protection against CSRF by adding uniqueness to the ViewState MAC as long as you set it to a new value for every session.
Note that ViewStateUserKey will not actually add a new value to the ViewState, but rather simply influence the MAC process so as to make every MAC unique per user session. Therefore, setting ViewStateUserKey to the current Session ID is acceptable.
Since Visual Studio 2012, the anti-CSRF mechanism has been improved.
The new strategy still uses the ViewState as the main entity for CSRF protection but also makes use of tokens (which you can generate as GUIDs) so that you can set the ViewStateUserKey to the token rather than the Session ID, and then validate it against the cookie.
Here's a blog post by Eric Johnson and James Jardine with an example of this implementation.
Considerations for AJAX
Depending on your application, you'll likely have to choose between using HTTP Headers or POST data to carry your anti-CSRF tokens.
Whatever you choose, the optimal validation method is indeed through tokens. This means you can follow the token strategy while creating either a custom header to hold the token value or just sending the token with the rest of the POST data.
Related Technical Impacts
Why refresh CSRF token per form request?
CSRF Prevention (official ASP.NET blog)
Preventing CSRF Attacks (official ASP.NET blog)
Anti-CSRF and Cookies
How to protect against CSRF by default in ASP.NET MVC 4?
How does ViewState protect against CSRF?
How To Fix CSRF using Microsoft .Net ViewStateUserKey and Double Submit Cookie, by Eric Johnson and James Jardine
CSRF Protection With Custom Headers (focus on the answer, not the question)
MSDN - Securing ViewState
MSDN - ViewStateUserKey
MSDN - HtmlHelper.AntiForgeryToken
MSDN - ValidateAntiForgeryTokenAttribute