Joe Stewart, director of malware research at SecureWorks, presented his survey at the RSA Conference, which opened Monday in San Francisco. The survey ranked the top 11 botnets that send spam; by extrapolating their size, Stewart estimated the bots on his list control just over a million machines and are capable of flooding the Internet with more than 100 billion spam messages every day.
The botnet at the top of the chart is Srizbi. According to Stewart, this botnet — which also goes by the names “Cbeplay” and “Exchanger” — has an estimated 315,000 bots and can blast out 60 billion messages a day.
While it may not have gotten the publicity that Storm has during the last year, it’s built around a much more substantial collection of hijacked computers, said Stewart. In comparison, Storm’s botnet counts just 85,000 machines, only 35,000 of which are set up to send spam. Storm, in fact, is No. 5 on Stewart’s list.
“Storm is pretty insignificant at this point,” said Stewart. “It got all this attention, so Microsoft added it to its malicious software detection tool [in September 2007], and that’s removed hundreds of thousands of compromised PCs from the botnet.”
The second-largest botnet is “Bobax,” which boasts an estimated 185,000 hacked systems in its collection. Able to spam approximately nine billion messages a day, Bobax has been around for some time, but recently has been in the news again, albeit under one of its several aliases.
Other researchers, notably those at a security start-up called Damballa, have been trumpeting a botnet dubbed “Kraken” — sometimes spelled “Kracken” — that they claim controls more then 400,000 computers. Stewart and others at SecureWorks believe Damballa has simply rebranded the older Bobax, which has several other nicknames besides Kraken, including “Bobic,” “Oderoor,” “Cotmonger” and Hacktool.Spammer.”
That mix-up over names is just one of the problems that Stewart hoped his research would solve, or at least reduce. “I’ve been covering botnets for a long time,” he said, “and there’s a lot of confusion about what botnets belong to what malware family. I want to try to shine some light on what malware belongs to what botnet, and what each botnet’s doing.”
To try to bring some organization to competing claims, often contradictory, of which botnets are on the rise and which on the skids, Stewart first “fingerprinted” each botnet. “There are enough differences to the SMTP ‘fingerprints’ for each botnet that we could separate them pretty accurately,” he said.
Individual bots implement the SMTP (Simply Mail Transfer Protocol) with minor variations, Stewart said. By developing network-based signatures, he was able to differentiate the collections.
He also estimated the size of each botnet by taking a one-day spam traffic sample from that bot — the sample derived from SecureWorks’ client base — and then using probabilistic counting methods, extrapolated to come up with a botnet total. Stewart said that past data collected from control server logs confirmed this estimating technique as “fairly accurate.”
The whole idea, he added, was to make it easier for everyone to keep track of the most dangerous botnets. “I hope this lets other researchers classify and track botnets better,” said Stewart. “Bobax, for instance, flew under the radar for over two years because of confusion. It was still around, but [anti-virus] vendors stopped recognizing [the malware].”
End users should get something out of his work, too. “I think it matters a lot to end users what a botnet’s called. They go to look for information, perhaps after they’ve been infected, and all they have is that it’s ‘agentxyz.'” But unless everyone is one the same page, that “agentxyz” may simply be a new alias. “Then they’d find hardly any information on what it is or what data it may be after. They won’t have a clear picture.
“I hope this trickles down to end users,” Stewart concluded.