Cross site scripting attack (Google example)

Source

The script (http://www.google.com/url?q=...) is normally used for redirecting the browser from Google's website to other sites.

For example, the following request will redirect the browser to http://www.watchfire.com:
http://www.google.com/url?q=http://www.watchfire.com 

When the parameter (q) is passed to the script with illegal format (The format seems to be: http://domain), a "403 Forbidden" page returns to the user, informing that the query was illegal. The parameter's value appears in the HTML returned to the user.

If http://www.google.com/url?q=USER_INPUT is requested, the text in the "403 Forbidden" response would be:
"Your client does not have permission to get URL /url?q=USER_INPUT from this server."

The server response lacks charset encoding enforcement, such as:
* Response headers: "Content-Type: text/html; charset=[encoding]".
* Response body: "<meta http-equiv="Content-Type" (...) charset=[encoding]/>".

Google's 404 NOT FOUND mechanism:
When requesting a page which doesn't exist under www.google.com, a 404 NOT FOUND response is returned to the user, with the original path requested.

If http://www.google.com/NOTFOUND is requested, the following text appears in the response:
"Not Found The requested URL /NOTFOUND was not found on this server."

The server response lacks charset encoding enforcement, such as:
* Response headers: "Content-Type: text/html; charset=[encoding]".
* Response body: "<meta http-equiv="Content-Type" (...) charset=[encoding]/>".

XSS vulnerabilities:
While the aforementioned mechanisms (URL redirection script, 404 NOT FOUND) escape common characters used for XSS, such as <> (triangular parenthesis) and apostrophes, it fails to handle hazardous UTF-7 encoded payloads.

Therefore, when sending an XSS attack payload, encoded in UTF-7, the payload will return in the response without being altered.

For the attack to succeed (script execution), the victims browser should treat the XSS payload as UTF-7.

IE charset encoding Auto-Selection:
If 'Encoding' is set to 'Auto-Select', and Internet-Explorer finds a UTF-7 string in the first 4096 characters of the response's body, it will set the charset encoding to UTF-7 automatically, unless a certain charset encoding is already enforced.

This automatic encoding selection feature makes it possible to mount UTF-7 XSS attacks on Google.com.

Solution:
Google solved the aforementioned issues at 01/12/2005, by using character encoding enforcement.