Every Twitter application you authorise has near-complete control over your account ? and that includes your private data
"What will we do," asked Bill Thompson recently, "when the next WikiLeaks hoard is every Twitter DM [direct message] or Facebook chat transcript?"
That's unlikely to happen, of course. Julian Assange's team are more concerned with international affairs than bothering the millions of casual Twitterers. If you're not on the public stage, the chances are that you're just not notable enough for Wikileaks.
Of course, there are still plenty of folks that would cause chaos in suburbia if they could, just for the laughs. I'm a little closer to this subject than most ? because I can read your private Twitter messages.
Well, maybe not yours, specifically. But the odds are good that ? if I wanted to ? I could look through the direct messages of several of the people who'll read this.
A few months ago, after Twitter had implemented its OAuth authorisation system, I was hired to put together a Twitter toy application ? one of those little web pages that offers some sort-of-interesting statistic about your tweets, as long as you log in with your Twitter account. Of course, it then invites you to tweet that statistic out to your friends ? along with a link and a small advertising message.
The trouble is that Twitter's authorisation process makes no distinction between small toys like that and big applications like TweetDeck that handle your entire account. Toys only need to read public messages and perhaps tweet once, but usually request, and are being given "read and write" permission, which means they can do every action Twitter can provide an authorised user: the power to change profile pictures, follow and block users, and ? crucially ? read direct messages. Changing your password doesn't lock them out either; you need to explicitly revoke their access.
So if I wanted to, I could use the authorisations given to the toy application to download the direct messages of all of its thousands of users. I'm not going to, of course ? it's illegal and unethical ? but curiosity is a powerful thing. Imagine a web page that simply asks "Whose inbox would you like to read?". Would you trust everyone you know not to use it? How about everyone they know? What if the attack had already been performed by someone else, and they'd actively leaked those messages to the world ? would you still be able to resist seeing what was in your friends' inboxes if the damage was already done?
Using existing applications' permissions isn't the only potential attack. Toy applications with poor security could provide a back door into people's Twitter accounts without their creators' knowledge. And those with espionage on their minds could create an application that works as advertised until it sees a particular user; the world's millions of inboxes might not be of interest, but specific ones might be.
Bill Thompson's WikiLeaks scenario, where every Twitter DM is somehow released, is unlikely. But without wanting to scaremonger, I'd say that unless Twitter starts using granular, Facebook-like authorisation, it's a matter of when ? not if ? an application goes rogue. Mischief isn't a strong motivator for releasing personal data (the risks are too high) but it only takes one script-kiddie cracker with a desire for notoriety, and suddenly The Pirate Bay is serving "two-million-twitter-DMs.rar" to anyone who's interested.
The bottom line is this: almost every Twitter application you authorise, no matter how trivial, has near-complete control over your account. This is not a new revelation, but it still takes a lot of people by surprise. If you haven't recently checked the Connections page of your Twitter account to see which applications you've authorised, you should. And revoke them unless you're certain you want to take the risk of them going rogue at some point.
? Tom Scott's web site is at tomscott.com; he's on Twitter at @tomscott
Source: http://www.guardian.co.uk/technology/blog/2010/dec/07/twitter-applications-direct-messages-privacy
No comments:
Post a Comment