For a long time I've been recommending companies to use Intrusion Detection Systems to detect infected/malicious machines on their own networks – instead of detecting inbound attacks (which will definitely happen, and the number of alerts will be in hundreds, if not thousands) they should detect outbound attacks. This way they can early detect potentially compromised internal machines when they phone home or download second stage binaries.

Matt Jonkman from Emerging Threats (http://www.emergingthreats.net) has been publishing Snort rules that detect non-standard User-Agent headers for a long time (if you're not using the ET rules set you're missing a lot!).

While this still catches a lot of malware, recently I saw more samples pretending to be Microsoft's BITS (Background Intelligent Transfer Service) – the service that's downloading all those updates Microsoft publishes every month.

As Microsoft made the BITS API available to anyone, malware authors can use that easily to download their own binaries, instead of Windows patches. Here's an example of a malware using BITS to download second stage binary – notice the target Host: header.



So, while monitoring User-Agent fields is still helpful in catching infected machines, we can expect that malware authors will use legitimate services such as BITS even more in the future. One thing we can possibly do is monitor such requests and when they contain the BITS User-Agent field check the Host: header; it should be easy to build a white list of allowed hosts.

If you have other ideas about how to improve User-Agent monitoring let us know.

--
Bojan