Researchers 'poison' Storm botnet

A group of German researchers has unveiled the first publicly released research attempting to actively disrupt a peer-to-peer botnet - using as their case study the notorious Storm worm.

The researchers were able not only to infiltrate Storm, gaining in the process the most precise estimates of its size to date, but also had success in disrupting its communications through a "poisoning" technique, according to the study.

The study, "Measurements and Mitigation of Peer-to-Peer-based Botnets," was authored by five researchers working with the University of Mannheim and Institut Eurécom, and was presented at the Usenix Workshop on Large-Scale Exploits and Emergent Threats (LEET) earlier this month in San Francisco.

Botnets have become among the most pressing threats on the internet, accounting for most spam and capable of directing large-scale denial-of-service and other categories of attacks.

[ See related stories where researchers  dispute Microsoft's claims to have crushed the Storm botnet, and security experts assert the botnet problem is worse than ever. ]

Most recently, botnets such as that created by the Storm worm have begun using peer-to-peer techniques for communications, eliminating the need for a central control server and making them far more difficult to shut down.

By taking a more active approach, the researchers found a way to "poison" the communications of the Storm bots, effectively disrupting them.

"Our strategy can be used as a way to disable the communication within the Storm botnet to a large extent," they wrote in the study. "As a side effect, we are able to estimate the size of the Storm botnet, in general a hard task."

Previous research has been based on passive techniques such as observing network events such as the number of spam emails thought to have originated from a particular botnet, the researchers said.

They said the new study is the first to use active techniques, crawling the P2P network, keeping track of all peers and distinguishing infected peers from benign ones based on behavior.

Crawling the "Stormnet" every 30 minutes from the beginning of December 2007 to the beginning of February 2008, the researchers found between 5,000 and 40,000 peers online at any given time, with a sharp increase in bots during the Christmas and New Years Eve periods.

The bots were located in more than 200 countries, with the biggest proportion in the U.S., at 23 percent.

The "poisoning" technique involves the keys used by Storm bots to establish communication. The researchers published a large amount of false content for particular keys.

"Our experiments show that by polluting all those hashes that we identified to be storm hashes, we can disrupt the communication of the botnet," the researchers wrote.

Another technique, called a "sybil" or "eclipse" attack, aiming to separate a part of the P2P network from the rest, proved ineffective.

Active interference with a botnet may carry serious legal consequences for researchers, since the botnet hosts are effectively computer systems belonging to third parties, who ordinarily are unaware that their systems are being misused.

In part because of this factor, previous research has focused on passive techniques for identifying the size and control structure of a botnet, leaving law enforcement authorities to take action, the researchers acknowledged.

The researchers said future efforts will focus on analyzing a second tier of systems that issue the actual commands, which might allow the identification of the operators of the Storm worm.

Techworld is an InfoWorld affiliate.