BO Team exploits Log4Shell in VMware vCenter and compromises a Russian software developer's CI/CD and code-signing server
Kaspersky ICS CERT documented a months-long intrusion by BO Team (aka Black Owl, Hoody Hyena, Lifting Zmiy) against an undisclosed Russian software developer. Kaspersky was contacted in June 2025. The group, assessed to be of Ukrainian origin, entered through an exposed SSH service on a Linux terminal server. It then exploited CVE-2021-44228 (Log4Shell) on a VMware vCenter 7.0.3 server to get root command execution. The attackers reached all parts of the IT infrastructure, including the GitLab source code, the Ansible system, the domain controller and the code-signing server. Kaspersky assesses that a supply chain attack was likely intended. Rapid response cut off their access before that could happen.
Entenda a falha
A vulnerabilidade
CVE-2021-44228, known as Log4Shell or LogJam, is a flaw in a Log4j component of VMware vCenter Server. In this intrusion the affected server ran version 7.0.3. BO Team used it to get a remote command line with root privileges on vCenter, then connected to the hypervisor over SSH. The articles do not say which vCenter builds are fixed or what exact conditions the attack needed, other than that the server was vulnerable and reachable from the compromised internal network.
Como funciona
The articles do not describe the internal mechanism of Log4Shell. What they describe is the sequence around it: the attackers first got into a Linux terminal server through an exposed SSH service and mapped the internal network. Their logins to the vCenter web interface failed, but they then targeted the vCenter server vulnerable to Log4Shell and obtained a root command line. From there they reached the hypervisor over SSH and planted a backdoor.
Por que é vulnerável
The articles give no root-cause analysis for Log4Shell. The relevant fact is that the victim's vCenter 7.0.3 server was still vulnerable to a flaw from 2021, so it had not been patched or mitigated. The articles do not describe what the fix changes.
Como foi descoberta
The flaw itself was not newly found. It is a well-known 2021 vulnerability that BO Team exploited. The intrusion came to light when the affected Russian software developer contacted Kaspersky ICS CERT in June 2025 about suspicious activity in its network. Kaspersky investigated, documented the months-long compromise and attributed it to BO Team (aka Black Owl, Hoody Hyena, Lifting Zmiy).
Explicação gerada por IA a partir da matéria e dos artigos, em 16/11/2026 05:16.
3 exploits públicos
Há código de exploração publicado no Exploit-DB para CVE-2021-44228. O código não é reproduzido aqui: o link leva à fonte, e a explicação foi escrita pela IA a partir do arquivo publicado.
-
AD Manager Plus 7122 - Remote Code Execution (RCE)
This entry describes Log4Shell (CVE-2021-44228) in ManageEngine ADManager Plus builds before 7122, reached through a request parameter on the product's search function in its web interface. The write-up needs only network access to the login page, with no valid credentials, and an unpatched version that still logs attacker-controlled input with a vulnerable Log4j. The authors first confirmed it with an out-of-band DNS callback, then claim full remote code execution on another target after evading a firewall and antivirus, but they publish no working exploit code. It is really a short research narrative and detection hint, not a ready exploit. Defenders should upgrade past 7122 and watch for JNDI lookup strings in web logs.
-
Apache Log4j 2 - Remote Code Execution (RCE)
This script exploits CVE-2021-44228 (Log4Shell) in Apache Log4j 2 versions 2.0-beta9 to 2.14.1, where a logged string with a JNDI lookup makes the application fetch and load remote code. The attacker needs no authentication, only a way to get text into something the target logs, and the target must be able to reach the attacker's server. The tool runs a rogue directory and web server that deliver a class opening a reverse shell, giving remote code execution. It is a working proof of concept, not verified by Exploit-DB.
-
Apache Log4j2 2.14.1 - Information Disclosure
This Python script targets Log4Shell (CVE-2021-44228) in Apache Log4j2 up to 2.14.1. It sends one HTTP request to a chosen web address with a malicious JNDI lookup string in the User-Agent header, so any vulnerable application that logs that header will make an outbound LDAP connection to the attacker's listener. The script runs a minimal fake LDAP server and captures the value of a Log4j lookup (by default the Java version, but environment variables can be chosen), leaking it back. No authentication is needed, only network reach to the target, a vulnerable version that logs the header, and outbound traffic allowed. It is a working information-disclosure proof of concept, not code execution.
Detalhes técnicos
- Initial access was an exposed SSH service on a Linux terminal server. The attackers used it for internal reconnaissance.
- After Log4Shell gave them root on vCenter, they pivoted to the hypervisor over SSH. There they planted a cron.hourly script named "pgcache" that downloads a second script from a nofrillsdns[.]com host and runs the GSocket utility.
- The backdoor is dormant while the DNS record points to 127.0.0.1. The attackers activate it by changing the record.
- They reused the hypervisor certificate to reach a second virtualization cluster.
- They stole secrets from VMs, which gave them GitLab repositories and the GitLab private key, and then the Ansible system.
- Credential theft used an lsass.exe dump through comsvcs.dll (a Nanodump attempt was blocked), registry dumps to decrypt ntds.dit, and Kerberos tickets. Shared administrator passwords across domains let them compromise the whole Windows estate.
- They modified /usr/bin/nxnode-login to capture credentials into /dev/shm/pulse-shm-3088612174.
- They brute-forced the terminal server administrator password with about 2,500 attempts.
- Defense evasion included timestomping, event log deletion, obfuscation and living-off-the-land tools.
- A Windows server used to sign builds was targeted. A WMI script and a backdoor were probably attempted there.
- In late July the attackers sent a phishing email, "Account disconnected from the mail server", which passed SPF and DKIM. This suggests the mail server's private key had been stolen. It was detected and blocked.
- Activity came from IPs 193.3.23[.]236 and 80.76.42[.]139, both at a Ukrainian provider.
Remediação
The article gives no explicit remediation. The implied steps are to patch Log4Shell on vCenter, avoid exposing SSH, use unique administrator passwords, protect code-signing servers, and monitor for the artifacts above.
Linha do tempo
-
BO Team exploits Log4Shell in VMware vCenter and compromises a Russian software developer's CI/CD and code-signing server
ler a análise desta fonte ocultar
Kaspersky ICS CERT documented a months-long intrusion by BO Team (aka Black Owl, Hoody Hyena, Lifting Zmiy) against an undisclosed Russian software developer. Kaspersky was contacted in June 2025. The group, assessed to be of Ukrainian origin, entered through an exposed SSH service on a Linux terminal server. It then exploited CVE-2021-44228 (Log4Shell) on a VMware vCenter 7.0.3 server to get root command execution. The attackers reached all parts of the IT infrastructure, including the GitLab source code, the Ansible system, the domain controller and the code-signing server. Kaspersky assesses that a supply chain attack was likely intended. Rapid response cut off their access before that could happen.
Technical Details
- Initial access was an exposed SSH service on a Linux terminal server. The attackers used it for internal reconnaissance.
- After Log4Shell gave them root on vCenter, they pivoted to the hypervisor over SSH. There they planted a cron.hourly script named "pgcache" that downloads a second script from a nofrillsdns[.]com host and runs the GSocket utility.
- The backdoor is dormant while the DNS record points to 127.0.0.1. The attackers activate it by changing the record.
- They reused the hypervisor certificate to reach a second virtualization cluster.
- They stole secrets from VMs, which gave them GitLab repositories and the GitLab private key, and then the Ansible system.
- Credential theft used an lsass.exe dump through comsvcs.dll (a Nanodump attempt was blocked), registry dumps to decrypt ntds.dit, and Kerberos tickets. Shared administrator passwords across domains let them compromise the whole Windows estate.
- They modified /usr/bin/nxnode-login to capture credentials into /dev/shm/pulse-shm-3088612174.
- They brute-forced the terminal server administrator password with about 2,500 attempts.
- Defense evasion included timestomping, event log deletion, obfuscation and living-off-the-land tools.
- A Windows server used to sign builds was targeted. A WMI script and a backdoor were probably attempted there.
- In late July the attackers sent a phishing email, "Account disconnected from the mail server", which passed SPF and DKIM. This suggests the mail server's private key had been stolen. It was detected and blocked.
- Activity came from IPs 193.3.23[.]236 and 80.76.42[.]139, both at a Ukrainian provider.
Remediation
The article gives no explicit remediation. The implied steps are to patch Log4Shell on vCenter, avoid exposing SSH, use unique administrator passwords, protect code-signing servers, and monitor for the artifacts above.