Hacker News new | ask | show | jobs
by MichaelAza 4653 days ago
It implies downloading a file onto the users machine without user consent which is, in itself, a problem. More importantly, an attacker could craft a torrent file that exploits vulnerabilities in the torrent client. If, just by visiting a site, an attacker can download an arbitrary file onto your machine and then have it automatically opened in a known program you're in big trouble.
2 comments

I don't understand, if the user is prompted to download the file using an external application it's no different than a direct download.

If users have their browsers configured to automatically start the download of any .torrent files without confirmation, twitter giving bogus .torrent is no more dangerous than $malware_site linking a .torrent. So that's not a security issue on twitter's site.

And anyway, I still fail to see how downloading a file (through bittorent or otherwise) constitutes a security breach on its own. Unless of course the bittorent client auto-executes binaries when it's done downloading, but that's just silly (and still nothing to do with twitter's security policy).

The flow of a (possible) attack is something like this:

1. User configures browser to automatically start torrent downloads when a ".torrent" link is clicked

2. User clicks twitt button which leads to a torrent file

3. The file is downloaded and opened in a torrent client

At this point, one could imagine a specifically crafted torrent file which exploits some vulnerability of the torrent client to gain (say) arbitrary code execution and now the user is, to use a mild term, screwed.

This attack could be used by any malicious site, really, but it's easier to get people to click a twitt button rather than some link on some site and besides, by preforming the attack this way the attacker would infect a sizable chunk of all internet sites (any site that uses the twitt button).

One could also imagine a specially crafted image file which exploits some vulnerability of the graphics library to gain arbitrary code execution. Then you just need the user to look at the twitter button.
True, though I'd think it would be easier to exploit a torrent client than a browser.
That attack vector has nothing to do with Twitter.
Did I imply it had something to do with twitter?

When this conjecture was posted I assumed someone hijacked a CDN used by twitter and used the twitt button as an attack vector by making it redirect to a torrent file.

I'm not saying twitter is trying to infect its users or something. In all probability, it's just a configuration screw-up and not an attack but (for all we know) it could be.

The torrent file it downloads is a binary, so it's most likely an auto-open exploit.