Fighting malware and spam

Riders on a Storm

Researchers hijack botnet - and find spam success rates lower than previously believed.

Less than 1 in 12 million spam emails sent through the infamous Storm botnet led to a purchase attempt, according to researchers at the University of California in San Diego and Berkeley - a much lower spam success rate than previously estimated.

To carry out their study, the researchers hijacked a part of the Storm botnet for 26 days, during which time they instructed it to send almost 350 million dummy spam emails advertising a dummy pharmacy site controlled by the researchers.

The researchers measured the activity of their campaign and concluded that less than one quarter of the emails sent had reached the recipients' mail clients, after which a large portion would have been filtered by spam filters. Users clicked on the link contained in the email in just over 0.003% of the emails and only 28 people (less than 1 in 12 million) attempted to buy something from the dummy website. The website deliberately gave an error just before the checking out process.

In two other spam campaigns the researchers mimicked a self-propagation attack, which attempted to install an executable on the user's machine. This method is commonly used by botnets such as Storm to replicate. The researchers found that this campaign had a higher success rate, with just over 1 in 229,000 users attempting to install the trojan.

The researchers believe that they controlled only about 1.5% of the total Storm botnet and, based on these assumptions, they estimated that the real botnet can generate revenue of between $7,000 and $9,500 a day. They estimated the number of new bots generated by self-propagation campaigns to be between 3,500 and 8,500 a day.

The ethics of a piece of research that involves actively sending spam may be questioned by security experts. However, the researchers claim to have effectively sidestepped legal and ethical issues: the campaigns under their control, they say, did not send any spam that would not otherwise have been sent, thus did not add to the total amount of harm done. Moreover, they did not collect people's personal data, nor did the self-propagation campaign lead to actual malware being installed on users' computers.

While the researchers admit that different spam tactics or different botnets might result in different metrics, the success rates they measured are significantly lower than those previously estimated by the anti-spam industry. With the retail price of spam thought to be slightly less than $80 per million emails, had this been a real spam campaign the spammers would have lost money. This led to speculation that Storm's botherders are involved in the pharmacy websites pushed by their spam.

More can be found at the BBC here or at The Register here, while the research paper can be downloaded in PDF format here.

11 November 2008

Tags: botnet, spam, storm.    del.icio.us   digg this

0 comments

Leave a comment

Poll

VB100 certification