IBM Security X-Force researchers studied the botnet activity of a malware variant that is used by cyber crime groups to illegally mine cryptocurrency. Examining two ShellBot botnets that appeared in attacks honeypots caught, the X-Force team was able to infect its own devices and become part of the live botnets, thereby gaining insight into how these botnets were managed internally. This post provides the details on IBM’s research and sheds light on the growing threat of cryptomining botnets to enterprise networks.
While seeing computational resources abused by cryptojacking operations is enough of a problem, the riskier consequence of the infection is that malware begets malware. A seemingly simplistic infection is still a foothold that can result in more sophisticated malware on the network down the line, which may end up exfiltrating data and even installing ransomware to extort the organization later on.
Cryptojacking Is the Name of ShellBot’s Game
With the exponential rise in the value of cryptocurrency, cybercrime endeavors based on these digital coins have been rising as well. Aside from the devastating rise of ransomware attacks, the illegal mining of cryptocurrency on devices one does not own, aka cryptojacking, has become a commercial grade threat used in the hands of lone criminals and organized groups alike. In some cases, cryptojacking operations that keep mining farms processing coins reached the magnitude of a $50 million business for their bot masters.
The ShellBot malware lives within this ecosystem. While it is a rather simple piece of Perl-based code, it enables attackers to mount internet relay chat (IRC)-controlled botnets that command coin mining on computers, Linux servers, Android devices and Internet of Things devices. The one requirement is having a weak password, as ShellBot’s typical entry point is a brute-force attack; the other is a command injection on servers that accept remote commands from the command-line interface (CLI).
While it started out as a basic IRC bot, over time ShellBot has been using effective exploits to compromise servers and devices. It started out with a ShellShock (CVE-2014-6271) campaign, which is how it got its name, but over the years has used Drupalgeddon (CVE-2018-7600) and other exploits that can compromise large swaths of devices. ShellBot has also been evolving its features to better spread through networks and disable competing infections to ensure all the computing power is used for its own goals. ShellBot’s objective, in most cases, is mining for Monero coin.
Brute-Forcing a Way In
ShellBot infections typically use brute-force attacks to guess the passwords of targeted servers and devices. In the botnets IBM X-Force examined, the most frequently used credential types helped identify the targets as misconfigured databases, FTP-servers, monitoring servers and other Linux machines.
By far the most frequent username for which the password was brute-forced was ‘root,’ followed by standard or default username strings such as ‘ubuntu’, ‘admin’, ‘user’ and ‘test’. There were also some standard database-type credentials such as ‘oracle’, ‘postgres’ or ‘mysql’.
Figure 1: The top username types used in ShellBot attacks (Source: IBM X-Force)
Botnet logs also showed the top passwords the malware managed to guess. Unfortunately for the targeted device owners, it was rather easy to figure out simple and default passwords.
Figure 2: The top weak passwords used in ShellBot attacks (Source: IBM X-Force)
ShellBot campaigns logged in IBM’s spam traps focused on verified ShellBot instances and ShellBot tactics, techniques and procedures (TTPs) that also launch a Perl-based payload. IBM X-Force saw quite a bit of activity every month, as shown in the figure below.
Figure 3: ShellBot activity by month, January to June 2021 (Source: IBM X-Force)
Infect and Control
ShellBot is dropped as a payload to systems and devices where a password was brute-forced. Immediately after a successful login, the infected machine/device receives a list of commands to execute; those include sending back system information, downloading and executing a PERL script, removing logs, removing the command history and deleting the payload itself.
Each ShellBot variant connected with a different botnet over an IRC channel. To get into the botnet, IBM infected some devices and followed the activity.
The ‘Blackcat’ Server
At the time of joining a server that was named Blackcat, the channel consisted of close to a hundred active bots and was about four months old. The channel itself was only used by the operator to address single or multiple machines, to which they would respond in a private chat.
Reconnaissance commands were regularly issued to the bots, with the goal of collecting information that would allow the botnet’s operators to zero in on valuable assets. The information the attackers looked for was root access, CPU and GPU information and system architecture (ARMV, etc.) Most notably, the Blackcat botnet operators were eager to find machines with NVIDIA GPUs, which is a graphic processing unit with higher compute power that translates into faster mining.
Once the server’s operators get the information on the system’s CPU/GPU, they group the bots in different channels accordingly.
Figure 4: Operators probing for GPU info and sorting bots into #nvidia channel
After sorting machines based on their mining capabilities, the operators would go on to download and execute a new ShellBot Perl script on those machines. The ‘pola’ version contained new IRC parameters, forcing the bots to join a different IRC server, which appeared like the operators were migrating active bots to a different network.
Figure 5: Operators deploying secondary malware with different IRC parameters to move bots to Pola server
Graduating to the Pola Server
The operation X-Force tracked was probably rather fresh. When the researchers followed the migrated bots to the Pola server, they noticed that it was created that very same day. It already contained 143 bots, which were most likely migrated from other channels. Within the next day, the number of bots doubled, suggesting that there must have been additional servers like Blackcat moving higher value bots to the Pola server.
Once on the Pola server, the filtering for higher value bots continued and a third payload — miner3.tgz — was deployed to the selected bots in a series of commands launched by the botnet’s operators.
The process begins with the removal of previous versions of the ShellBot malware and those of other cryptominers that might be resident on the machine/device. Once the machine is clean of potential rivals, the operators go on to install their own miner. An archive is then unpacked into a .cache directory, in which the script .x is executed. This launches XMRig as well as XHide (process hider) on the infected machine.
Figure 6: Cryptominer download
#Armv – the DDoS Channel
Botnets of all types, and especially those that command infected devices, are often used for distributed denial-of-service (DDoS) attacks. With a simple command, operators order the bots to browse to the target website and attempt to flood it with traffic and cause a denial of service. The botnets had a channel for bots that took part in DDoS attacks. The floods lasted 500 seconds each and did not appear very powerful. It is possible that the channel had a few low-value devices on it or was used as a testing ground.
OpSec by Segmentation
From the way bots are filtered, moved to other botnets and run new malware versions, it can be inferred that there is some sort of tiered segmentation.
Conversations found on the server are proof that trading bots among operators are rather common, which explains how botnets can double in size quickly. The benefit of tiered segmentation is that it makes it very hard to find IRC server information on Tier 2 servers and higher, since these are only deployed by Tier 1 operators. The only way to join a Tier 3 botnet would be to be moved up the chain and follow the newly deployed Perl scripts. Along the way, it becomes much easier for the different botnet operators to filter out research bots, increasing the security of the operation.
ShellBot is publicly available code, so it is harder to attribute it to any one group. Often, botnets using widespread code can be used by anyone, but they can see more meaningful activity from specific actors and groups.
Links to Romanian-Speaking Operators
What IBM found throughout its research, especially during the infiltration of IRC servers, are clues to an operation being managed by Romanian-speaking bot masters. For one, X-Force found Perl scripts stating flood.ro as the author, and some instances of ShellBot were hosted on a Romanian news site. The Perl scripts further linked to Romanian-speakers’ IRC channels on multiple servers. One of the operators used the Romanian hostname blackcat.ro on IRC. All threat actors conversing in the channels were fluent in Romanian. These links to Romania line up with threat intelligence on ShellBot by other security researchers in the past few years.
Links to the Outlaw Gang
The Romanian threat group tracked as Outlaw, which was identified in 2018, has been observed to use ShellBot to target different organizations. After comparing data from IBM’s own research with TTPs from previously reported Outlaw attacks, X-Force found similarities in the threat actors’ TTPs to those in the Blackcat and Pola servers.
A screenshot of the Perl script had previously appeared in an article by TrendMicro and is identical to one of the samples used for the infections in the cases IBM saw. Another identical script is the Pola script applied to bots that graduated the Blackcat server.
A malware analysis conducted by Yoroi on a 2020 Outlaw campaign mentions the same bash scripts, run and upd, which X-Force found in the cryptominers distributed on the Pola server. Lastly, TrendMicro published an article on Outlaw campaigns, which discovered the use of an old process hider, XHide, to mask the mining process XMrig. The hash of the XHide binaries, h32 and h64, contained in the cryptominer matched up to the ones used in one of the Outlaw campaigns.
Is this evidence that the servers IBM infiltrated were associated with the Outlaw group itself? It’s possible, and it’s also plausible that other threat actors in this space copied the TTPs or bought the scripts from someone else.
Mitigating the Risk of Cryptojacking Malware
Cryptojacking malware can be an insidious and long-term attack that’s often hard to detect but can be damaging in a number of ways: impaired server performance, accumulation of electricity costs and overheating devices, to name a few. What puts companies at greater risk is the residence of malware on networked assets and devices, which can allow attackers to strengthen their foothold down the line. That risk can become any other type of attack given the time and the motivation of the attackers.
In order to minimize the risk of a ShellBot infection, it is vital to properly configure all public-facing devices with strong credentials and ensure that logging is in place. For servers that can be accessed remotely, it is wise to add multifactor authentication and disable the option to run CLI commands remotely if that is a requirement the business needs. In addition, all outbound IRC traffic should be monitored or entirely blocked, as this may be an indicator of a ShellBot infection and potential data exfiltration that is not related to business needs. Network system monitoring should be implemented to detect excessive resource utilization and users should be educated about the risk and signs of cryptojacking on their devices.
Indicators of Compromise
- Blackcat: 184.108.40.206
- Same operator as on Blackcat: 220.127.116.11, 18.104.22.168
- Pola: 22.214.171.124
Pola script download:
Recent Shellbot attacks detected on IBM’s honeypots:
- Attacker IPs (brute-forcing):
- 126.96.36.199 14/06/2021
- 188.8.131.52 12/06/2021
- 184.108.40.206 02/06/2021
- 220.127.116.11 16/05/2021
- 18.104.22.168 15/042021
- Shellbot download URLs