Five Sinkholes of newGOZ
By Dennis Schwarz and Dave Loftus
It has been a few weeks since news broke of the Zeus Gameover variant known as newGOZ. As has been reported, the major change in this version is the removal of the P2P command and control (C2) component in favor of a new domain generation algorithm (DGA).
The DGA uses the current date and a randomly selected starting seed to create a domain name. If the domain doesn’t pan out, the seed is incremented and the process is repeated. We’re aware of two configurations of this DGA which differ in two ways: the number of maximum domains to try (1000 and 10,000) and a hardcoded value used (0x35190501 and 0x52e645).
Date based domain generation algorithms make for excellent sinkholing targets due to their predictability, and provides security researchers the ability to estimate the size of botnets that use them. With this in mind, we have gathered five days worth of newGOZ sinkhole data. Our domains are based on the first configuration, since this configuration seems to be used the most in the wild.
As with all sinkhole data, many variables can affect the accuracy of victims such as network topology (NAT and DHCP), timing, and other security researchers. However, we feel that the data provides a good estimation of the current scope of this new threat.
Monday, July 14
Four days after the discovery of newGOZ, our first sinkhole saw 127 victims. To corroborate our initial data set, SecureWorks reported seeing 177 victims connect to their sinkhole a few days earlier on July 11.
Friday, July 18
An 89% increase to 241 victims.
Monday, July 21
Over the weekend we saw a 78% increase to 429 victims, mostly in the eastern half of the United States.
Friday, July 25
As reported by Malcovery Security on July 22, they saw a large spam campaign distributing newGOZ by the Cutwail botnet. This campaign appears to have been very successful. On July 25, we saw an 1879% increase to 8494 victims—the rest of the United States is covered.
Monday, July 29
Over the weekend and 19 days after its discovery, our fifth and final sinkhole for this post saw a 27% decrease to 6173 victims. This is most likely due to victims cleaning themselves up from that last spam campaign. Latin America, South Africa, South East Asia, and New Zealand start filling in.
In aggregate and over three weeks, our five sinkholes saw 12,353 unique source IPs from all corners of the globe:
The most infected country was the United States followed by India. The top 10 were:
In addition, a number of organization types were affected, the top being:
Pondering on the five days worth of newGOZ sinkhole data above, some thoughts come to mind:
First, will the threat actor continue to use the same DGA configuration that they’ve been using so far? Empirically, there seems to be more security research sinkholes populating the DGA namespace than actual C2 servers. There is also the second DGA configuration that hasn’t received much use yet. Additionally, as we’ve seen, the actor is willing to completely replace the C2 mechanism altogether.
Second, will the botnet continue to grow and at what rate? The sinkhole data for July 25 suggests that the second Cutwail spam campaign was relatively successful. Will future waves continue this trend?
Finally, with the infection numbers at a fraction of what they were in the P2P version of Zeus Gameover, how long will the threat actor focus on rebuilding their botnet before they return to focusing on stealing money?