< Back to all blog posts

Mozi-ing around C2s with Black Lotus Labs

The GreyNoise Team

GreyNoise sees a lot of botnet activity, both benign and malicious, through our fleet of global sensors. We enlisted Black Lotus Labs®, the threat research arm of Lumen, to help us bring you more information about these botnets and their Command and Control (C2) hosts. A botnet’s C2, as the name implies, is a host from which bots receive their commands, download malicious files, and/or simply report back. Effectively, the C2 is the brain of the operation and without one, a bot may simply sit dormant.

Black Lotus Labs' mission is to find and disrupt C2s to make the internet a cleaner and safer place. The amount of noise generated by these botnets, driven by their C2s, is astounding. For example, check out the traffic for May just for the Mozi botnet:

Figure 1: Volume of netflow records  associated with known Mozi botnet IPs observed  by Black Lotus Labs for the month of May, 2021.
Figure 1: Volume of netflow records associated with known Mozi botnet IPs observed by Black Lotus Labs for the month of May, 2021.

Some reports estimate that bots generate 25% of all internet traffic. This reflects what we see at GreyNoise - for example, on any given day we tag around 30k IPs as ‘Mirai’ or ‘Mirai Variant’, one of the most prevalent bots in the wild. We are always looking for ways to identify sources of internet noise, so hunting botnets and their C2s with the Black Lotus Labs team was a natural fit.

Identifying C2s Using Graph Analysis

To kick off our project, Black Lotus Labs enriched their netflow data using IPs tagged as ‘Mirai Variant’ in GreyNoise and applied some basic graph algorithms to identify a list of potential C2 IPs. These algorithms mapped 23 suspected C2 IPs in Black Lotus Labs’ data that communicated the most with several hundred GreyNoise ‘Mirai Variant’-tagged IPs. This one-to-many relationship is indicative of a traditional C2. Some of the potential C2 IPs in Black Lotus Labs’ data were previously identified Mirai C2s but, interestingly enough, others were identified as a botnet family known as Mozi, a newer family that eschews the traditional C2 model.

Figure 2: Potential C2 determined by a one-to-many relationship to tagged bots.
Figure 2: Potential C2 determined by a one-to-many relationship to tagged bots.

Mozi exhibits many of the characteristics of Mirai with one major exception: it has no central C2. Instead, Mozi is a peer-to-peer (P2P) botnet where every infected host is both a bot and a C2. Each peer propagates configurations and hosts payloads while also performing bot duties such as participating in DDoS, scanning the internet, and exploiting hosts to expand the botnet. For more on Mozi and P2P botnet technology, check out Black Lotus Labs’ analysis.

Figure 3: Centralized botnet (left) vs P2P botnet (right)
Figure 3: Centralized botnet (left) vs P2P botnet (right)

Identifying C2s Using Request Analysis

From the GreyNoise side of the project, we decided to look at request payloads within scanner traffic to see if we could identify C2s. We observed that, despite their difference in centralization, botnets like Mirai and Mozi are both notorious for inserting C2 addresses (IPs or domains) into their initial exploit attempt. Typically these exploits will execute a script that fetches a malicious payload from the C2 address and initializes the bot.

If you’ve ever looked at traffic hitting your network perimeter, you have probably seen a request like this:

Figure 4: Example request with C2 IP.
Figure 4: Example request with C2 IP.

If you extract and check the IP, you might discover, unsurprisingly, that the host is a C2. That’s it. No advanced analytics. No machine learning. No blockchain. Just IP and domain extraction.

We decided to leverage this insight, and test if this was an accurate way to identify P2P Mozi family C2s - that is, IPs that scan like a bot, AND deliver peer-to-peer C2 addresses. Our approach was to extract a set of IPs matching this request pattern from our data, and then compare the results with Black Lotus Labs C2 data.

Using this method, we compiled a list of 3,368 suspected C2 IPs that appeared to be delivering requests with embedded C2 addresses. Our free Analysis tool confirmed that 97% of these IPs scanned the internet within the last 90 days . So our hypothesis was that this combination of bot and C2 behaviors allows us to accurately identify P2P Mozi family C2s.

Figure 5: Potential C2s observed scanning by the GreyNoise Analyzer.
Figure 5: Potential C2s observed scanning by the GreyNoise Analyzer.

To test the hypothesis, we asked Black Lotus Labs to analyze the IP list, and identify any C2s already known to them. They found that our list contained 962 IPs previously identified as C2s or botnet peers.

Figure 6: Left, percentage of IPs analyzed by Black Lotus Labs . Right, breakdown C2 families for the analyzed IPs.
Figure 6: Left, percentage of IPs analyzed by Black Lotus Labs . Right, breakdown C2 families for the analyzed IPs.

In total, 28% of our potential Mozi IPs were identified by Black Lotus Labs as C2s. Of those, 98% were confirmed as Mozi. So while this is a promising start at identifying suspected C2 IPs, it doesn’t provide conclusive evidence that IPs exhibiting this behavior belong to the Mozi family. Further research is required to profile the remaining 71%, which are most likely simple bots.

Identifying Vulnerabilities Using Extracted C2s

Why is it important to identify C2 IPs like Mozi? Using the confirmed C2 data in hand, we found we can now pivot around the addresses (both IPs and domains) to help identify the vulnerabilities being exploited. For example, take the following C2 domain:

bp65pce2vsk7wpvy2fyehel25ovw4v7nve3lknwzta7gtiuy6jm7l4yd[.]onion[.]ws

We found that more than 50% of traffic containing this C2 domain belonged to IPs, probably bots, exploiting the same three vulnerabilities: Terramaster TOS (CVE-2020-28188), Zend Framework (CVE-2021-3007), and Liferay Portal (CVE-2020-7961).

The FreakOut botnet is known to exploit this unholy trinity. Although we already have tags for all three of these vulnerabilities, this demonstrates how we can use C2 addresses to automate the process of identifying and tagging known unknowns: vulnerabilities used by botnets.

Recall that bot traffic comprises almost a quarter of all internet traffic. Developing and expanding these techniques allow us to closely examine some of the most common noise on the internet. Any vulnerability checks or exploit used by a botnet like Mirai or Mozi is bound to be one of the most well used on the internet. By knowing botnets, we know noise.

Additionally, we want to refine and share these fun C2 addresses, like cnc[.]tacobelllover[.]tk, with our customers and community as a data feed for your projects, research, and work. If this interests you, create a GreyNoise account and join our Community Slack to give us feedback.

Insights