Slashdot Mirror
Cross Site Scripting Discovered in Google
Security Test writes "Yair Amit posted a message early this morning to The Web Security Mailing List outlining a Cross Site Scripting flaw in Google that allows an attacker to carry out Phishing Attacks."
If there ever was an endorsement for web-based applications, this is it. When a bug is fixed in Windows or Linux, it stays active in the wild for months or years because many users don't update. With web apps the user basically gets an "update" each time they visit the site. If Google fixed the problem on December 1, the vulnerability could have been announced the same day without any kind of negative impact.
The downside is that this only works if the app provider is a proprietary vendor with a closed architecture. If 3rd parties are allowed to create extensions or if users can create their own utilities/add-ons then centralized patching would likely introduce the same types of incompatibilities and breakages that current OS patches can introduce. Worse, centralized control might mean that users have no choice but to live with the patched version.
Two wrongs don't make a right, but three lefts do.
I'm always blown away by how the Internet security market works and self-correct itself without any regulation.
A major web site has a flaw. White hat and black hat "hackers" find that flaw, exploit it, and either abuse it or let the web site know about it. The web programmers go in and close the exploit because it affects how their customers use the service and could open them up to some liability.
This is the way the free market works. I'm a huge fan of how quickly the Internet (anthropomorphically) adapts to the changing needs of the billion of users. Some exploits that aren't fixed by the owners of code are fixed by third parties -- sometimes for profit and sometimes for free. Before we can even write one law to attempt to solve problems, others are already attacking the problems.
I'd like to see it stay this way. Every time we move forward to create legislation to protect the end user (see CAN-SPAM and a myriad of other laws), we see failure time and again. The loopholes in the laws make them irrelevant quickly, and all we get out of that is wasted money and wasted time.
Let the growth and expansion occur freely. We'll see some bad times (new viruses and new spam exploits) but we'll see those fixed in short order. If they don't get fixed, why is the Internet still chugging along and growing every day?
This is reported as a Google.com bug, which is partially true. But this is only one half of the problem. The other half of the problem (mentioned in the full article) is due to a dubious feature in Internet Explorer: when it gets a page without a specified character encoding, it does not rely on default values for the encoding (which should be iso-8859-1 for HTML or UTF-8 for XHTML).
Instead, Internet Exploerer tries to guess the encoding of the contents by looking at the first 4096 bytes of the page and checking the non-ASCII characters. In the case of the cross-site scripting attack decribed here, the problem is that IE would silently set the encoding of a page to UTF-7 in case some characters in the first 4096 bytes looked like UTF-7. This silent conversion to UTF-7 by Internet Explorer in a text that Google assumed to use the default encoding allowed the attackers to bypass the way Google was filtering "dangerous" characters in some URLs.
The article puts the full blame for the vulnerability on Google.com. I think that a part of the blame should also be shared by the Internet Explorer designers (and any other browser that does unexpected things while trying to guess what the user "really meant").