This article is more than 1 year old

Unholy trinity of flaws put Google users at risk

Doomwatchers count the ways

If you use Google to send email, organize photos or help administer your website, doomwatchers have cataloged three new ways to steal your data and compromise the security of your users. All three of the techniques rely on cross site scripting, or XSS, in which hackers inject unauthorized code by making it appear as if it's hosted by a trusted website.

The most serious vulnerability resided in the so-called polls application, a part of Google Groups. It made it possible to steal contacts and messages from Gmail accounts. A Google spokesman on Monday afternoon said the flaw had been fixed.

Multiple pieces of proof-of-concept code posted online graphically demonstrated the potential for attacks that target the weakness. One stole all contacts listed in a Gmail account, while a second sent all incoming Gmail messages to an email account of the researcher's choosing.

"If you're good at JavaScript, writing a good exploit for those vulns is [a] trivial matter" Giorgio Maone, a researcher told El Reg. He added that the weaknesses could be exploited "serially," meaning that a single piece of attack code will compromise virtually any Gmail account.

The exploit code worked on at least four browsers, including Internet Explorer, Firefox, Opera and Konqueror, at least when they are using default settings and are not running extensions that block the running of javascript. To be pwned, a victim had to be logged in to Gmail while opening the attack website.

A second vulnerability plagues the Google search appliance, which Google sells so webmasters can do scraping and other types of internal searches. By creating a specially crafted URL, an attacker can inject code or overwrite pages of a third-party site that uses the appliance.

Attack scenarios include the stealing of cookies used to log in to the third-party site or the alternation of a trusted site so it prompts an unsuspecting user for personal information and then transmits it to the attacker.

The disclosure came in a post on the Mustlive blog, which included examples of URLs that were said to cause two sites - one run by ICANN and the other by The University of York to steal a user's cookies. (We plugged both addresses into version 7 of IE and only the latter appeared to work, so it's possible that ICANN has already plugged the hole.)

According to this Google search, as many as 200,000 sites were vulnerable to the attack at time of writing.

A third vulnerability could be used to steal photos designated in Google's Picasa picture organizer simply by luring a user to a malicious website. It is a highly complex exploit that marries a variety of techniques, including XSS, cross application request forgery, Flash and URI handler weakness exploitation.

A Google spokesman said the company had recently become aware of the flaw in its search appliance and would offer more details about a fix after a more thorough investigation. Google officials are unaware of the vulnerability being targeted in the wild.

The spokesman also said the company has made it more difficult to exploit the Picasa weakness by educating users about the consequences of installing new buttons to the application. The button, which can now only be installed after clicking OK on a confirmation window, must be installed for the attack to work, he said.

"Google takes security issues very seriously and will respond swiftly to fix verifiable security issues," the spokesman said. "Some of our products are complex and take time to update."

This unholy trinity of serious weaknesses is reminiscent to a week-long stretch in late May and early June in which four Google vulnerabilities were documented. Like those discovered this time, most involved XSS errors, which are often the result of trying to make multiple services scattered across a website work seamlessly.

Other sites to be bitten recently by the XSS bug include Yahoo! and TJMaxx.

While Google's security team usually score high marks for policing the site and safeguarding users from scams, the mass of vulnerabilities are a reminder that as code becomes increasingly sophisticated, even the most elite developers make mistakes that can lead to security breaches.

"It's pretty endemic of where the web applications security community is mentally," said Robert Hansen, CEO of secTheory.com, who publishes on the ha.ckers blog under the moniker of RSnake. "They're looking to make a point, which is the companies that do the most to protect themselves are probably the most vulnerable. Even spending hundreds of millions of dollars on this still has not gotten them to the point where they're secure." ®

More about

TIP US OFF

Send us news


Other stories you might like