Boffins devise early-warning bot spotter
Conficker's Achilles Heel
By Dan Goodin • In Security • At 20:13 GMT 5th November 2010
Researchers have devised a way to easily detect internet names generated by so-called domain-fluxing botnets, a method that could provide a first-alarm system of sorts that alerts admins of infections on their networks.
Botnets including Conficker, Kraken and Torpig use domain fluxing to make it harder for security researchers to disrupt command and control channels. Malware instructs infected machines to report to dozens, or even tens of thousands, of algorithmically generated domains each day to find out if new instructions or updates are available. The botnet operators need to own only a few of the addresses in order to stay in control of the zombies. White hats effectively must own all of them.
It's a clever architecture, but it has an Achilles Heel: The botnet-generated domain names – which include names such as joftvvtvmx.org, ejfjyd.mooo.com, and mnkzof.dyndns.org – exhibit tell-tale signs they were picked by an algorithm rather than a human being. By analyzing DNS, or domain name system, traffic on a network, the method can quickly pinpoint and disrupt infections.
“In this regards, our proposed methodology can point to the presence of bots within a network and the network administrator can disconnect bots from their C&C server by filtering out DNS queries to such algorithmically generated domain names,” the researchers wrote in a paper that was presented this week at the ACM Internet Measurement Conference in Australia.
The method uses techniques from signal detection theory and statistical learning to detect domain names generated from a variety of algorithms, including those based on pseudo-random strings, dictionary-based words, and words that are pronounceable but not in any dictionary. It has a 100-percent detection rate with no false positives when 500 domains are generated per top-level domain. When 50 domains are mapped to the same TLD, the 100-percent detection rate remains, but false positives jump to 15 percent.
The technique was developed by Sandeep Yadav, Ashwath K.K. Reddy, and A.L. Narasimha Reddy of Texas A&M's Electrical and Computer Engineering department, and Supranamaya Ranjan of Sunnyvale, California-based Narus. A PDF of their paper is here. ®