[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Android Browser Cross-Application Scripting (CVE-2011-2357)

 Android Browser Cross-Application Scripting (CVE-2011-2357)

1) Background
Android applications are executed in a sandbox environment, to ensure that no
application can access sensitive information held by another, without adequate
privileges. For example, Android's browser application holds sensitive
information such as cookies, cache and history, and this cannot be accessed by
third-party apps. An Android app may request specific privileges during its
installation; if granted by the user, the app's capabilities are extended.
Intents are used by Android apps for intercommunication. These objects can be
broadcast, passed to the startActivity call (when an application starts another
activity), or passed to the startService call (when an application starts a
service). Normally, when startActivity is called, the target activity's
onCreate method is executed. However, under AndroidManifest.xml it is possible
to define different launch attributes, which affect this behavior. One example
is the singleTask launch attribute, which makes the activity act as a
singleton. This affects the startActivity call: if the activity has already
been started when the call is made, the activity's onNewIntent member function
is called instead of its onCreate method. Moreover, if the target activity is
not in focus when the call is made, Android automatically inserts the
FLAG_ACTIVITY_BROUGHT_TO_FRONT flag to the input Intent, which it doesn't do

2) Browser Internals
The Android browser's main activity, as defined in its manifest file, is
BrowserActivity. This is defined with the singleTask launch mode. The input
Intent for the activity may hold a URL, which is opened and then rendered by
the browser.
* The activity's onCreate member function, tries to restore the
  browser's previous state. If it fails to do so, it creates a new tab, with the
  input Intent's URL (if there is one), or else with the defined homepage.
* The activity's onNewIntent member function, has the following characteristic:
  If the Intent is not a search Intent (for example, if its action is
  ACTION_VIEW), or if it is a search Intent with a query string defined in URL
  form, then it performs a resolution in order to deduce which tab to load the
  given URL under (again, if there is no input URL, the homepage is used as a
  * If the intent contains FLAG_ACTIVITY_BROUGHT_TO_FRONT flag, it tries to
    find a tab with a matching application ID (as indicated by the Intent's
    Browser.EXTRA_APPLICATION_ID extra string) or with a matching URL. If it
    fails to do so, it loads the URL in a new tab, as long as the number of
    opened tabs is less than MAX_TABS (usually 8). Otherwise, it opens the URL
    in the current tab.
  * As a last resort, it loads the URL in the current tab.

The Browser app uses the WebView class as the underlying engine. If the WebView
class has already loaded a URL, and the same instance is used to load a
javascript:// URI, then the javascript is executed in the domain of the loaded
URL. This is the desired behavior, as it allows applications to inject scripts
into loaded pages, and control the WebView. However, this means that the
browser must take special care if it reuses the same WebView instance, in order
to avoid a Cross-Application Scripting vulnerability.

3) Vulnerability
A 3rd party application may exploit Android's Browser URL loading process in
order to inject JavaScript code into an arbitrary domain thus break Android's
sandboxing. There are two vectors that can achieve this:
1. The malicious application causes the Android's browser to reach the MAX_TAB
   limit. From then on URLs are loaded under the current tab. The attacking
   application can open MAX_TAB URLs by calling startActivity <MAX_TAB> times
   with the attacked domain. On the <MAX_TAB+1>th call, the attacking app can
   insert a javascript:// URI, which will be opened in the context of the
   attacked domain. It should be denoted that the sent Intent should be
   combined with the FLAG_ACTIVITY_BROUGHT_TO_FRONT flag because it is likely
   that the Browser will have UI focus from the second intent and forward, in
   which case Android won't attach this flag automatically and the crucial code
   fragment under onNewIntent will not be executed.
2. Sending two consecutive startActivity calls. The first call includes the
   attacked domain, and causes Android's browser to load it. The second call
   contains the javascript code. If the time interval between the two intents
   is small enough, then it is likely that the browser will have UI focus when
   the second startActivity call is made, therefore the input intent won't have
   the FLAG_ACTIVITY_BROUGHT_TO_FRONT flag and, as explained in the previous
   vector, the JavaScript URI will be opened under the current tab, i.e. the
   attacked domain.

4) Impact
By exploiting this vulnerability a malicious, non-privileged application may
inject JavaScript code into the context of any domain; therefore, this
vulnerability has the same implications as global XSS, albeit from an installed
application rather than another website. Additionally, an application may
install itself as a service, in order to inject JavaScript code from time to
time into the currently opened tab, thus completely intercepting the user's
browsing experience.

5) Proof-of-Concept
The following is a PoC for the second technique:

public class CasExploit extends Activity
   static final String mPackage = "com.android.browser";
   static final String mClass = "BrowserActivity";
   static final String mUrl = "http://target.domain/";;
   static final String mJavascript = "alert(document.cookie)";
   static final int mSleep = 15000;

   public void onCreate(Bundle savedInstanceState) {
         try {
         catch (InterruptedException e) {}
         startBrowserActivity("javascript:" + mJavascript);

   private void startBrowserActivity(String url) {
      Intent res = new Intent("android.intent.action.VIEW");
      res.setComponent(new ComponentName(mPackage,mPackage+"."+mClass));

6) Vulnerable versions
Android 2.3.4 and Android 3.1 have been found vulnerable.

7) Vendor Response
Android 2.3.5 and 3.2 have been released, which incorporate a fix for this bug.

The fixes can be found in the following commits:
* http://android.git.kernel.org/?p=platform/packages/apps/Browser.git;a=commit;h=afa4ab1e4c1d645e34bd408ce04cadfd2e5dae1e
* http://android.git.kernel.org/?p=platform/packages/apps/Browser.git;a=commit;h=096bae248453abe83cbb2e5a2c744bd62cdb620b

Patches are available for Android 2.2.* and will be released at a later date.
Organizations can contact security@xxxxxxxxxxx for patch information.

Android has communicated information about this vulnerability to their
partners, and all new Android compatible devices are required to
incorporate this bug fix:
* http://source.android.com/compatibility/overview.html
* http://source.android.com/faqs.html#compatibility
* http://android.git.kernel.org/?p=platform/cts.git;a=commit;h=7e48fb87d48d27e65942b53b7918288c8d740e17

Android Market actively scans all Android Market applications to detect and
prevent exploitation of security vulnerabilities.

8) Additional information
* Original advisory:
* Demo of PoC:       http://www.youtube.com/watch?v=BzUpbcrWufs

9) Credit
* Roee Hay <roeeh@xxxxxxxxxx> of IBM Rational Application Security
Research Group
* Yair Amit <yairam@xxxxxxxxx>

10) Acknowledgments
We would like to thank the Android Security Team for the efficient and quick
way in which they handled this security issue.