Is that malware found on your client machine the sign of a targeted attack or a routine bot-herding run? How do you know for sure?

Botnet hunters from Damballa are using some traditional network monitoring techniques to determine the size and scope of botnets -- information that can even help distinguish between a direct attack or a random bot recruitment.

“We are working on ways to better [calculate] the numbers of these botnets with some accuracy,” says Christopher Davis, director of threat analysis for Damballa. Davis and Damballa chief scientist and co-founder David Dagon will discuss their company’s botnet research techniques at Black Hat D.C. next week.

Damballa researchers basically reverse-engineer the malware code that arrives at one of their customer’s client machines, and then study how it communicates with its command and control (C&C) server. Then, using a DNS cache-inspection technique, combined with tracking the C&C server’s IP packet identifier in TCP/IP, they can take more accurate counts of the number of bots, C&C servers, and the potential scope of a particular botnet.

Damballa is basically putting a new spin on some existing techniques. Phishers, for example, have been known to use DNS cache inspection for reconnaissance before staging an exploit on an organization. “Although these techniques have been around and known for a long time, we’ve never heard of anyone applying them to botnet research,” Davis says.

These methods are useful for bot malware that provides only limited visibility into the botnet’s inner workings, such as HTTP-based botnets. An IRC-based botnet -- or even Storm, which uses peer-to-peer communications -- wouldn’t require these techniques because they are more transparent and simpler to track, Davis says.

Learning about the size of the botnet behind a piece of malware can provide some clues as to whether it’s a targeted attack or a bot-recruiting run. “You’re never able to say 100 percent that this wasn’t a targeted attack. But generally, the bad guy isn’t likely to have a [targeted] enterprise join a 20,000-member botnet. It’s going to be a small one,” Davis says.

With DNS cache inspection, the researchers query regional servers for the “bad” domain. “We can then see if this bad domain has been requested there…it wouldn’t be in the cache if no one else had asked for it,” Davis says.

So if an Atlanta-based ISP’s DNS server retains the bad guy’s domain in its cache, it’s likely that users in that area are getting infected, he says. This gets them closer to learning about the range of the bot population.

Then comes the so-called IPID technique. Davis says the researchers can determine how many packets it takes for the malware’s update or download, so they can use that count to calculate roughly how many bots there are. IPID basically provides a packet counter they can refer to in each query to the C&C server. “Watching IPID, we can tell when [the C&C server] sends a command or someone downloads the malware from it,” Davis says.

The downside of these methods, however, is that they’re relatively “noisy,” Davis says, which could potentially blow the researchers’ cover and prompt the bad guys to relocate their servers. So the researchers try to keep a low profile when scanning DNS servers and when sending packets to the botnet C&C servers so their presence isn’t detected, he says. The goal is to be able to better determine growth, spread, and shifts in a botnet.

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.

Damballa Inc.