The Network Security Lab at 360 provided evidence that the Mirai botnet has a built-in domain generator algorithm (DGA) feature. In this blog, we share some of what we learned about frequent traffic patterns of Mirai DGA domains.

In brief:

  • Mirai DGA domains may see up to 10K queries per hour.
  • A frequent pattern is seen where Mirai DGA domains initially appear within a day, then re-emerge in a sustained period of query volume.
  • We began to see Mirai DGA query volumes spike after December 2 (although the .pw variant appears to have shown up a couple days earlier).
  • On December 8, Mirai hard-coded C2 domains displayed a coordinated spike in query volume.

A quick note: these conclusions were generated from our unique traffic and may be different from conclusions drawn by others performing a similar analysis.

CLIENT IP – DOMAIN GRAPH

To find these DGA domains, we pivoted off known hardcoded C2 domains used by Mirai:

  • zugzwang.me
  • tr069.support
  • tr069.online
  • tr069.tech

In addition, we found another tr069 with a different TLD:

  • tr069.pw

Leveraging our distributed bipartite Client IP-Domain graph we found client traffic which overlapped these hard-coded domains with wordless domains – the DGA domains.

From mining this graph, we were able to find DGA domains with highly similar traffic patterns.

In fact, in a later analysis we found a majority of these domains (at one time) hosted on the IP:  93.190.142.201 which we first saw on December 5 at 05:54 UTC.

SAMPLED DOMAINS

Here’s a sample of 4 top-level domains (TLDs) with what appear to be Mirai associated domains – tech, support, online, and pw. More specifically, we will discuss query volumes from this sample of domains:

  • lvfjcwwobycj.tech
  • vmdefmnsndoj.tech
  • nympompksmfx.tech
  • qjqubpciajoc.tech
  • oornsduuwjli.tech
  • exvdaajegjur.support
  • xpknpxmywqsr.support
  • bpmsfckfkrpr.support
  • bwhrdaumwuvn.support
  • xpknpxmywqsrhe.online
  • kedbuffigfjs.online
  • srrys.pw
  • binpt.pw
  • kciap.pw
  • mziep.pw

QUERY VOLUME VERSUS CLIENT VOUME

The query volume is a raw count of the total number of processed DNS requests for a domain, which may include repeat visitors. In this analysis we group queries by hour. This is in contrast to a related count, the client volume, which tracks unique visitors to a domain – similarly, grouped by hour.

For example, here’s an example of the unique number of clients querying the Mirai C2 domain zugzwang.me over the last 7 days in the USA:

zug_clients

FIGURE 1: Unique USA client volume querying zugzwang.me in the last 7 days.

Similarly, here’s the USA client volume to tr069.pw in the last 7 days:

tr069pw_clients

FIGURE 2:Unique USA client volume querying tr069.pw  in the last 7 days.

Notice, the number of clients within the USA querying these domains varies: one domain had a maximum within one hour around 15 clients, while the other 2 clients.

In contrast, the plots below show the total query volume for individual domains grouped by hour.

REFERENCE DOMAINS

For reference, you might wish to know the traffic patterns of the C2 domains mentioned above. The query volumes are taken from aggregating all the queries processed in our resolvers around the world.

screen-shot-2016-12-11-at-10-00-50-pm

FIGURE 3: zugzwang.me

screen-shot-2016-12-11-at-10-01-26-pm

FIGURE 4: tr069.support

screen-shot-2016-12-11-at-10-02-20-pm

FIGURE 5: tr069.online

screen-shot-2016-12-11-at-10-03-00-pm

FIGURE 6: tr069.tech

screen-shot-2016-12-11-at-10-25-47-pm

FIGURE 7: tr069.pw

It appears that on December 8, Mirai hard-coded C2 domains displayed a coordinated spike in query volume. Next, we’ll look at query volumes from a sample of Mirai DGA domains.

TLD: tech

The first TLD we look at is tech. We show 5 different domains with what appears to be 4 different traffic patterns.

screen-shot-2016-12-11-at-9-39-16-pm

FIGURE 8: lvfjcwwobycj.tech

screen-shot-2016-12-11-at-9-40-40-pm

FIGURE 9: vmdefmnsndoj.tech

screen-shot-2016-12-11-at-9-41-18-pm

FIGURE 10: nympompksmfx.tech

screen-shot-2016-12-11-at-9-42-01-pm

FIGURE 11: qjqubpciajoc.tech

screen-shot-2016-12-11-at-9-42-34-pm

FIGURE 12: oornsduuwjli.tech

TLD: support

The second TLD we look at is support. We show 4 different domains with what appears to be 2 different traffic patterns. We note, these traffic patterns overlap with those of tech.

screen-shot-2016-12-11-at-9-43-24-pm

FIGURE 13: exvdaajegjur.support

screen-shot-2016-12-11-at-9-44-09-pm

FIGURE 14: xpknpxmywqsr.support

screen-shot-2016-12-11-at-9-44-41-pm

FIGURE 15: bpmsfckfkrpr.support

screen-shot-2016-12-11-at-9-45-24-pm

FIGURE 16: bwhrdaumwuvn.support

TLD: online

The third TLD we look at is online. We show 2 different domains, each with a different traffic pattern. We note, these traffic patterns overlap with those of tech & support to some extent.
screen-shot-2016-12-11-at-9-45-58-pm

FIGURE 17: xpknpxmywqsrhe.online

screen-shot-2016-12-11-at-9-46-30-pm

FIGURE 18: kedbuffigfjs.online

 

TLD: pw

The last TLD we look at is pw. We show 4 different domains, each with a different traffic pattern. We note, these traffic patterns appear like some of those of techsupport, and online.

We point out that srrys.pw appears earlier than December 2, 2016 and binpt.pw has more sustained traffic than previously discussed domains.screen-shot-2016-12-11-at-9-47-43-pm

FIGURE 20: srrys.pw

screen-shot-2016-12-11-at-9-48-31-pm

FIGURE 21: binpt.pw

screen-shot-2016-12-11-at-9-49-08-pm

FIGURE 22: kciap.pw

screen-shot-2016-12-11-at-9-49-36-pm

FIGURE 23: mziep.pw

 

Notes:

We also found a couple other domains which may interest other researchers. Since these domains don’t appear to be generated by the same DGA, they don’t appear above. The two other domains are:

screen-shot-2016-12-11-at-9-50-51-pm

FIGURE 24: mufoscam.org

screen-shot-2016-12-11-at-9-46-59-pm

FIGURE 19: novotele.online

This post is categorized in: