The flaw lies in Firefoxs Password Manager software, which can be tricked into sending password information to an attackers Web site.

For this attack to work, attackers need to be able to create HTML (Hypertext Markup Language) forms on the Web site, something that is allowed on blogging and social networking sites.

The attack was used in a MySpace phishing attack reported in late October. In that attack, users registered a MySpace account named login_home_index_html and used it to host a fake log-in page that exploited the flaw.

This page sent MySpace username and password information to another Web site, and MySpace users who visited the page using Firefox could have easily had their information compromised.

Firefox developers rate this bug critical, according to an entry in the projects Bugzilla database.

The flaw arises because Firefoxs Password Manager does not perform a thorough enough check when it is deciding whether to send password information, and then does not ensure that password information is being sent to the server that requested it. In the MySpace attack, for example, Firefox would check to see if the form was coming from the MySpace.com domain, but did not make sure that the password information was being sent back to a MySpace server.

From a programming point of view, this is almost like a typo. Ironically I think that is why it has not been discovered until now. It was just way too obvious.

Microsoft Internet Explorer (IE) is also susceptible to these type of attacks because, like Firefox, it does not ensure that password information is being sent to the same server that requests it.

But IE is less likely to be tricked because it does a more thorough job in checking to see where the log-in form is coming from before it automatically submits password and user information.