Tuesday, September 29, 2009

Smaller botnets dominate the enterprise network

I've been a little quiet on the blog these last couple of weeks - having spent quite a bit of time either writing or delivering new threat presentations (3 last week alone). Last week while I was in Miami speaking at Hacker Halted, a colleague (Erik Wu) was in Geneva for VB2009 presenting our latest findings of a study of some 600 different botnets encountered within enterprise networks.

I finally got around to pulling a quick blog together for the Damballa site covering one of the findings - related to the size of botnets. You can find a copy of the posting Botnet Size within the Enterprise on the Damballa blog and cross-posted below.

One additional thing I'd like to point out though... the number of hosts compromised which are members of small botnets is still only a fraction of the total number of botnet members found within the enterprise - i.e. we're talking about botnets operated by 600 botnet masters, rather the 1m+ compromised hosts we studied.

Cross-posting begins...

Last week at the VB2009 conference in Geneva, Erik Wu of Damballa presented some of our latest research findings. There’s been quite a bit of interest in these botnet findings – largely because very few people have had the opportunity to examine enterprise-focused botnets, rather than the noisy mainstream Internet botnets – in particular the differences between the two types of networks. So, with that in mind, I wanted to take some time here to provide more information about the key findings (I’ll try to cover other aspects in later blogs).

While we often observe plenty of stats pertaining to just how big some of the largest Internet-based botnets are (reaching in to the tens-of-millions), the spectrum of Enterprise-botnets appear to be different – at least from Damballa’s observations across our enterprise customers.

Based upon Damballa’s observations of some 600 different botnets encountered and examined within global enterprise businesses over three months, we found that small (sub 100 member) botnets account for 57 percent of all botnets.

Biggest Botnets within Enterprise

Fig 1. Biggest Botnets within Enterprise

As you can see in the pie chart above, Huge botnets (10,001+ members) accounted for 5 percent, Big botnets (501-10,000) accounted for 17 percent, Average botnets (101-500) accounted for 21 percent and Small (1-100) reached 51 percent of the 600 different botnets found successfully operating within enterprise environments.

The average size of the 600 botnets we examined hovered in the 101-500 range on a daily basis. Why do I use the term “on a daily basis”? Because the number of active members within each botnet tend to change daily – based upon factors such as whether the compromised hosts were turned on or part of the enterprise network (e.g. laptops), whether or not they had been remediated, and whether or not the remote botnet master was interactively controlling them.

While many people focus on the biggest botnets circulating around the Internet, it appears that the smaller botnets are not only more prevalent within real-life enterprise environments, but that they’re also doing different things. And, in most cases, those “different things” are more dangerous since they’re more specific to the enterprise environment they’re operating within.

Taking a closer look at all these small botnets (sub 100 victim counts), we noticed that the vast majority of them are utilizing many of the popular DIY malware construction kits out there on the Internet. These DIY kits (such as Zeus, Poison Ivy, etc.) normally retail for a few hundred dollars – but can often be downloaded for free from popular hacking forums, pirate torrent feeds and newsgroups – and are usable by anyone who knows how to use an Internet search engine and has ever installed software on a PC before.

It looks to me as though these small botnets are highly-targeted at particular enterprises (or enterprise vertical sector), typically requiring a sizable degree of familiarity of the breached enterprise itself. I suspect that in some cases we’re probably seeing the handy-work of employees effectively backdooring critical systems so that they can “remotely manage” the compromised assets and avoid antivirus detection – similar to the problems enterprise organizations used to have with people placing modems in machines for out-of-hours support. The problem though is that the majority of these “freely available” DIY malware construction kits are similarly backdoored. Therefore any employee using these free kits to remotely manage their network are also providing a parallel path for the DIY kit providers to access those very same systems – as evidenced with these small botnets often having multiple functional command and control channels.

As for the other small botnets, it looks like these are more professionally managed – with botnet masters specifically targeting corporate systems and data within the victim enterprise. These small botnets aren’t being used for noisy attacks (such as those seen throughout the Internet concerning spam, DDoS and click-fraud) – but rather they’re often passively monitoring the enterprise network to identify key assets or users and then going for high value items that can be either used directly (e.g. financial controller authentication details for large money transfers) or high value salable data (e.g. extracting copies of customer databases and source code to applications). Unfortunately for their enterprise victims, the egress traffic is almost always encrypted – so the only way of finding out specifically what information has been leeched away is going to rely upon detailed forensics and log analysis of the compromised hosts and the systems they interacted with.

The net result is that these smallest botnets efficiently evade detection and closure by staying below the security radar and relying upon botnet masters that have a good understanding of how the enterprise functions internally. As such, they’re probably the most damaging to the enterprise in the longterm.

– Gunter Ollmann, VP Research

No comments:

Post a Comment