The university uses so-called darknet sensors that were set up about six years ago in order to keep track of malicious activity. With funding from the U.S. Department of Homeland Security, computer scientists have banded together to share data collected from sensors around the world place sensors around the world.
“The goal is to get close enough so you can actually start mapping out how the spread started,” said Jon Oberheide, a graduate student with the University of Michigan who is working on the project.
That’s not an easy job. To find the minuscule clues that will identify the victim, researchers must sift through more than 50 terabytes of data, hoping to find the telltale signatures of a Conficker scan.
One of the ways that Conficker moves about is by scanning the network for other vulnerable computers, but it can be really hard to spot it for certain, Oberheide said. “The hard thing is to find the exact Conficker scanning activity, because there is a lot of other scanning going on,” he said.
Tracking down patient zero has been done, however. In 2005, researchers tracked the 2004 Witty worm’s first victim, (pdf) a U.S. military base, and even identified the European IP address used to launch the attack.
It’s been years since anything as widespread as Conficker has surfaced however, so there have not been many chances to reproduce this effort.
When Conficker first appeared in October, though, researchers caught a break. Other worms had dodged this kind of analysis by blocking the darknet IP addresses, but Conficker’s authors didn’t do that. “We were kind of surprised that it did this completely random scan, and didn’t blacklist our particular sensors,” Oberheide said. “If they’d done a little bit of research, they could have discovered our [network].”
Soon after the Conficker outbreak the Michigan researchers saw a big spike on their sensors, which they attributed to the worm. The network was collecting about 2G of data per hour in November, but these days it’s closer to 8G. “The increase in activity we’ve seen on these Darknet sensors is… incredible,” Oberheide said. “Now this data is actually useful; we can go back six months and see what this worm was actually doing,” he added.
Another group, called CAIDA (the Cooperative Association for Internet Data Analysis) publisheda Conficker analysis earlier this month. The Michigan researchers hope to post a similar analysis of their data in with the next few weeks, but it could be months before they narrow things down to patient zero.
In the meantime, “the goal is to get close enough so you can actually start mapping out how the spread started,” Oberheide said.