Stop Counting Bots

How many bots are on the Internet, and why should we care? This is an argument I've been making since the late 1990s, and it is high time I got it in writing outside of closed circles.Ever since botnets became a mainstream concern -- years after they were a real threat in the wild -- our industry has been trying to count them. Most recently I came across a number citing more than 100 million bots on the Internet.

That number is irrelevant as much as it is untrue.

To count bots, we first need to define what a bot is. Is it the instance of the Trojan horse that responds to remote commands? Or is it the computer that may have 20 or 30 different active bots on it, making it an effective member of 20 or 30 different botnets?

Maybe a bot is an equivalent for an infected IP address, which changes constantly on dynamic IP ranges (for home users) just like it can include thousands on thousands of infected computers behind NATs and other private networks.

For the purpose of this post, I will define a bot as an instance of an active member in a botnet, no matter how many other types of malware may be running on the computer. I will also define an infected computer as an IP address that for the past 24 hours has been active in any number of botnets, no matter how many instances of malware are on that host.

Now considering the supply pool for bots is virtually unlimited -- pragmatically there is an unlimited number of computers vulnerable to infection -- and some criminals like to collect infected hosts, these days many botnets in the field are smaller and only used as they are needed.

In 2003-4, I made this argument and mentioned that if we have enough bots to destroy the Internet 10,000 times over, then does it really matter if the number of bots is 8,000 or 12,000? The issue was that while my argument is solid, in 2005 policy-makers were slowly becoming interested in botnets, and numbers impress people. Further, this would be the first question they'd ask.

In 2004-5 when we were discussing these issues, a good guy named David Dagon mentioned we should not be counting bots, but rather botnets. At the time this was solid advice, but it no longer holds true, either.

Today, botnets are an on-demand product and can operate in many different ways. Counting them does not really help with measuring the threat. What needs to be measured is the criminals themselves, the scale of their operations, and what they are using the botnets for.

Botnets are just an infrastructure from which most criminal activity on the Internet today operates -- just like in information warfare, when such botnets are hired or taken over to participate in large-scale assaults.

It's time we stopped counting weapons and started collecting intelligence on the wielders in a more in-depth fashion than we have so far, and by entities other than in the security industry.

Follow Gadi Evron on Twitter: http://twitter.com/gadievron.

Gadi Evron is an independent security strategist based in Israel. Special to Dark Reading.