2009 has been a notable year for malware and malicious online activity for a number of reasons and several of them relate to what is known as botnets. A zombie, or a bot, is a PC infected by malware that brings it under the remote control of a criminal. Criminals run networks that can range from thousands to millions of infected machines and they use them to power most of the cybercrime we see today including spam, DDoS, scareware, phishing, and malicious or illegal website hosting. They have a finger in every cybercriminal pie.
In the first half of the year, the Conficker worm (also known as Downadup or Kido) stole all the headlines in the malware world. Eventually the Conficker botnet was seen to deliver standard cybercriminal payloads, such as spambots and Fake AV (or scareware), much to the disappointment of some of the more hysterical commentators. Just because the outbreak received so much coverage that died away just as rapidly, don’t be fooled into thinking this threat has gone away. The Conficker Working Group, an alliance of security vendors, researchers and other commercial organisations is currently showing around 6 million unique IP addresses as appearing to be infected with this malware.
An unrelated, but important trend in 2009 was the exponential increase in the abuse of social networking providers for malicious purposes. The enormous active user populations on sites like Facebook, Twitter and MySpace prove a very attractive lure to organised online crime and its attendant money-making, bot recruitment and Fake AV pushing scams. Facebook has been abused by rogue Apps, designed to fool users into clicking links that reward the creator through pay-per-click affiliate advertising networks. It has also been used to spread malware through many means; malicious links in wall posts and messages, malware designed specifically to hijack accounts and by external compromise of legitimate Facebook Apps. The Koobface family of malware (also a botnet) has evolved over the course of 2009; it was initially spread through malicious messages and wall posts with links to fake YouTube sites punting a supposed codec in order to view the video. The codec of course was nothing of the sort and led to infection and account hijacking. Koobface now though has evolved to the point where it is fully capable of creating its own fake Facebook profile pages, complete with confirmed Gmail address, photo and biographical data. These fake accounts then set about joining networks and sending friend requests again all in a completely automated fashion.
Here’s where it gets interesting, in addition to spamming and malware, web 2.0 sites have been abused in new and concerning ways over the course of 2009. Twitter and Google Reader have been used as the landing page in spam campaigns, to attempt to overcome URL filtering in email messages. In recent months Twitter, Facebook, Pastebin, Google Groups and a Google AppEngine have all been used as surrogate Command & Control servers for botnets, and just last week it was reported that a Zeus botnet was leveraging compromised servers inside Amazon’s EC2 cloud for command and contro. These public forums have been configured to issue obfuscated commands to globally distributed botnets, these commands often contain further URLs which the bot then accesses to download commands or components.
The attraction with these sites and services lies in the fact that they offer a public, open, scalable, highly-available and relatively anonymous means of maintaining a command and control infrastructure, which at the same time further reduces the chance of detection by traditional technologies. Whilst network content inspection solutions could reasonably be expected to pick up on compromised endpoints that are communicating with known-bad sites (command & control servers), or over suspicious or unwanted channels such as IRC; it has been historically safe to assume that a PC making a standard HTTP GET request, over port 80 to a content provider such as Facebook, Google or Twitter, even several times every day, is as acting entirely normally. However, as botnet owners and criminal outfits seek to further dissipate their command and control infrastructure and blend into the general white noise of the internet, that is no longer the case.
It is no coincidence that much the innovation in 2009 has been around command & control systems for botnets. The vast majority of old-school IRC controlled botnets are shut down within 24 hours and peer-to-peer bots often leave visible signatures too, leading to their neutralisation at machine level. One factor of web 2.0 botnet controls that I would expect cybercriminals to be currently evaluating is the single point of failure represented by relying on a single provider such as Facebook or Google–shut down the malicious Facebook page and you disable the botnet. Botnet creators have invested significant amounts of time and code in distributing their management infrastructure, in fast-flux and in peer-to-peer protocols. We can fully expect them to carry these lessons learned into the newer “cloud-enabled” botnet. It is entirely possible that the capability of the latest Koobface variant to create multiple automated profiles could be leveraged to mitigate against the single point of failure inherent in using a single Facebook or Twitter profile as a covert channel.
When it comes to botnets it would be really nice to be able to say “it’s getting better”. It’s not. More and more computers are being infected, and they are staying infected for longer.