One of the best ways to avoid malware is for virtual machine technology to scan for any suspicious-looking Web domains and make sure they’re walled off, but the latest version of the DGA Changer downloader may render such sandboxes nearly useless.

Researchers from Seculert, which first discovered DGA Changer two years ago, recently published a blog post showing how quickly the malware has evolved. Now, when cybercriminals who use the downloader encounter a virtual machine or sandbox, DGA Changer can generate a series of domain names that look real enough to evade detection.

As if that weren’t enough, SecurityWeek reported that attackers have actually purchased some of those fake domains so that, if an organization were to test them, it would encounter a dummy executable file that makes it even harder to keep the real malware at bay. This isn’t the first variant of DGA Changer that researchers have spotted in the wild this year, but if it works as experts believe, it could quickly become one of the most effective.

Although these are still early days for DGA-type threats, research is already underway to try to prevent such malware from getting past enterprise firewalls. The Register profiled an early prototype of a detection technique underway at Cisco, which compiles possible domain names and then compares them against various sources, scoring them for potential security threats. Similarly, researchers at Carnegie Mellon University have created classification algorithms to identify malware tools such as DGA Changer.

Of course, if domains aren’t mere gibberish, this kind of ranking analysis will be increasingly difficult to do. In fact, Threatpost suggested that, like it or not, it may be up to human beings to look at domain logs and second-guess those that aren’t familiar if they want to ensure the greatest level of protection. As machine learning matures, it’s possible that the industry will also come up with ways to better automate such work. But as this latest variant of DGA Changer proves, it will take a combination of both technology and human ingenuity to predict how this latest brand of cybercriminals will act — and to get in front of them before they come up with their next trick.