Facebook OAuth flaw allows gaining full control over any Facebook account

Facebook OAuth is used to communicate between Applications & Facebook users, to grant additional permissions to your favorite apps. To make this possible, users have to 'allow or accept' the application request so that app can access your account information with required permissions.

As a normal Facebook user we always think that it is better than entering your Facebook credentials, we can just allow specific permissions to an app in order to make it work with your account.

Today whitehat Hacker 'Nir Goldshlager' reported 'The Hacker News' that he discovered a very critical vulnerability in Facebook's OAuth system, that allowed him to get full control over any Facebook account easily even without 'allow or accept' options.

For this purpose he hunt the flaw in a very mannered way i.e

Step 1) Understanding the OAuth URL

Step 2) Finding a way to use custom parameters in URL

Step 3) Bypassing OAuth 'Allow' button request at user end

Goldshlager explained everything on his blog via a post, so will also go step by step:

1.) Understanding the OAuth URL

The Facebook OAuth dialog URL is something like shown below:

https://www.facebook.com/dialog/oauth/?app_id=YOUR_APP_ID&next=YOUR_REDIRECT_URL&state=YOUR_STATE_VALUE&scope=COMMA_SEPARATED_LIST_OF_PERMISSION_NAMES

Where app_id is the application ID and next parameter must contains the URL of the respective app domain only. For example app_id=2389801228 belongs to 'Texas Holdem Poker' app, So the 'next' parameter will allow only zynga.com domain (i.e next=https://zynga.com), otherwise Facebook will block that action.

2.) Finding a way to use custom parameters in URL

Goldshlager found that Facebook was allowing him to use facebook's sub domain in next parameter in the URL ie. https://beta.facebook.com/#xxx!/messages/. But '#xxx!' was not working for all browsers. After fuzzing the URL characters, he found that %23~! and %23%09! worked for all browsers.

This finding was enough to redirect user to any file or folder at Facebook domain.

Then he developed a simple Facebook application (i.e touch.facebook.com/apps/testestestte) ,which was just to redirecting users to remote site (i.e. files.nirgoldshlager.com) with access token, where a log file was ready to store all access tokens.

3.) Bypassing OAuth 'Allow' button request at user end

Till now attacker was able to redirect user to a fake app, which was passing victim's access tokens to a 3rd party domain where attacker was logging access tokens. But the main issue was still there i.e without user interaction, app will not work. That means, one have to click 'allow' button as shown below.

So, to bypass this, he discovered that there are many built-in Facebook applications i.e 'Facebook Messenger app' that can access full permissions (read inbox, outbox, manage pages, manage ads,access to private photos, videos, etc.) from the victim's account without user interaction i.e no need to click 'allow' button.