BO Team Weaponizes Log4Shell to Turn VMware Infrastructure Into a Supply Chain Launchpad
Overview
BO Team used compromised Linux and virtualization infrastructure to move through a Russian software developer’s environment, steal credentials, establish persistent access, and reach systems responsible for source code, infrastructure automation, and commercial software signing. The investigation began in June 2025, while the associated research was published on November 5, 2026. Therefore, the publication date should not be interpreted as the start of a new November 2026 intrusion.
The available forensic evidence indicates that exploitation of CVE-2021-44228, commonly known as Log4Shell, is the most likely explanation for root-level command execution observed on a vulnerable VMware vCenter Server 7.0.3 deployment. Investigators reproduced exploitation against the same vCenter build, but the original exploit request could not be conclusively recovered because complete historical evidence was unavailable.
BO Team subsequently reached hypervisors, virtual machines, source-code repositories, the Ansible orchestration environment, and a server used to sign compiled commercial software. Investigators suspected that the attackers may have intended to inject malicious code into a product before signing or steal signing material for use with malicious files. However, no successful modification or distribution of signed software was confirmed.
The intrusion therefore created a serious software supply-chain risk. Access to development, automation, and code-signing systems could have allowed BO Team to compromise trusted software operations and extend the attack from the original developer to its customers.
Background
BO Team has been associated with disruptive and intelligence-gathering activity involving information theft, ransomware deployment, public claims of compromise, and attacks against organizations of strategic interest. Its operations combine legitimate administrative utilities, publicly available security tools, custom backdoors, credential theft, and extensive attempts to hide evidence.
In this incident, the first reliably identified compromise involved an internet-accessible Linux terminal server. The attackers authenticated through SSH using a service account for which username-and-password authentication remained enabled. From that foothold, BO Team scanned internal networks, investigated available services, and moved toward the organization’s VMware virtualization environment.
The organization’s development environment increased the potential impact. Access to virtualization, GitLab, Ansible, software-signing infrastructure, and Windows domain systems could provide visibility into source code, deployment secrets, administrative credentials, and systems capable of distributing trusted software changes across a wider estate.
Vulnerability Details
| CVE ID | CVSS Score | EPSS | CISA KEV | Affected Product | Vulnerability Type |
|---|---|---|---|---|---|
| CVE-2021-44228 | 10.0 Critical | Approximately 100% | Yes | Apache Log4j Core versions 2.0-beta9 to before 2.3.1, versions 2.4 to before 2.12.2, and versions 2.13.0 to before 2.15.0. The investigated environment used VMware vCenter Server 7.0.3 with a vulnerable Log4j component. | JNDI lookup injection enabling unauthenticated remote code execution under vulnerable configurations. |
Apache identifies the precise affected Log4j Core ranges as
[2.0-beta9, 2.3.1),
[2.4, 2.12.2), and
[2.13.0, 2.15.0).
The corresponding security-fixed releases are Log4j 2.3.1 for Java 6, 2.12.2 for Java 7, and 2.15.0 for Java 8 and later.
Log4Shell allows attacker-controlled lookup strings to cause vulnerable Log4j installations to retrieve and process content from an external location. When the vulnerable logging path is reachable, exploitation may lead to arbitrary code execution with the privileges of the affected Java process.
In this incident, root-level commands appeared on the vCenter appliance after unsuccessful web-interface sign-in attempts and without a corresponding SSH session. The presence of a vulnerable Log4j component, the observed command syntax, and successful laboratory reproduction make Log4Shell the leading explanation. It remains a high-confidence forensic assessment rather than a conclusively recorded exploit request.
The Log4j 2.15.0 correction was later found to be incomplete for separate conditions addressed under CVE-2021-45046. Organizations should consequently deploy the latest supported Log4j and VMware releases instead of treating these historical minimum versions as the preferred current remediation.
Visual Attack Flow
BO Team authenticates to an internet-accessible Linux terminal server using a service account with password-based SSH access.
The attackers scan internal address ranges and investigate SMB, Kerberos, SSH, HTTP, HTTPS, WinRM, LDAP, and other reachable services.
Forensic evidence and laboratory reproduction indicate that CVE-2021-44228 was the most likely source of root-level execution on VMware vCenter. The original exploit request was not conclusively recovered.
BO Team reaches ESXi infrastructure, installs disguised backdoors, alters timestamps, and retrieves additional commands and remote-access tooling.
Compromised certificates, credentials, and private keys enable movement into additional clusters, virtual machines, GitLab, and Ansible systems.
BO Team reaches a server used to sign compiled commercial software. Investigators suspect an intended supply-chain operation, but no successful malicious software distribution is confirmed.
Access to source code, signing infrastructure, orchestration systems, virtualization platforms, and Windows domains creates opportunities for information theft, ransomware, and trusted software compromise.
Attack Methodology
Phase 1: Initial Access Through Exposed SSH
The first reliably identified incident event was the compromise of an internet-accessible Linux terminal server. BO Team authenticated using a service account for which username-and-password authentication remained enabled.
Phase 2: Internal Service Discovery
From the terminal server, BO Team scanned internal IP ranges for SMB, Kerberos, SSH, Telnet, FTP, HTTP, HTTPS, WinRM, and LDAP services. Recovered activity also indicated the use of an SSH proxy and web-security tooling to investigate internal portals.
Phase 3: Likely Log4Shell Exploitation of VMware vCenter
After unsuccessful attempts to sign in through the VMware vCenter web interface, root-level command processes appeared without a corresponding SSH session. A Base64-encoded command established a reverse shell to another terminal server.
The affected vCenter Server 7.0.3 deployment had a vulnerable Log4j component loaded. The command syntax resembled previously documented exploitation behavior, and investigators reproduced Log4Shell exploitation against the same vCenter build. These findings make CVE-2021-44228 the most likely cause of the observed root-level execution, although the original exploit request was not conclusively recovered.
Phase 4: Hypervisor Persistence and Remote Access
BO Team accessed /root/.ssh/id_rsa, connected to the hypervisor over SSH, and placed a backdoor named pgcache under /etc/cron.hourly/. The filename imitated a PostgreSQL component, while modified timestamps made the script appear older.
The script contacted attacker-controlled infrastructure and retrieved additional commands. One payload installed remote-access functionality that allowed the attackers to maintain interactive control.
Phase 5: Expansion Across Virtual Infrastructure
The attackers abused credentials and certificates obtained from compromised infrastructure to connect to another virtualization cluster. They collected information about virtual machines and firewall rules, created virtual network interfaces, and executed commands inside guest systems from the hypervisor.
Phase 6: GitLab and Ansible Compromise
BO Team searched compromised virtual machines for secrets and reached the organization’s GitLab server. Access to source-code repositories and the server’s private key enabled further movement into the Ansible orchestration environment.
Phase 7: Code-Signing Server Compromise
BO Team reached a Windows server used to sign compiled builds of the organization’s commercial software. Investigators recovered evidence of blocked attempts to inject malicious code on this system and found that relevant security-event records had been deleted or overwritten.
Access to this server could have allowed the attackers to insert malicious code before a legitimate build was signed or steal digital-signing material for use with malware. Combined with BO Team’s interest in source-code repositories and software-composition analysis, this activity supported the suspected supply-chain objective.
The intrusion was detected and contained before investigators confirmed that malicious code had been incorporated into, signed with, or distributed through the organization’s commercial software.
Phase 8: Windows Credential Theft
Credentials stored in service scripts helped BO Team approach the Windows domain infrastructure. The attackers attempted to dump lsass.exe and later used legitimate operating-system components to produce memory dumps.
BO Team also copied registry data and used it with ntds.dit to recover domain credentials. Kerberos tickets, reused administrative passwords, SSH, RDP, SCP, and legitimate utilities supported movement through multiple domains.
Phase 9: Defense Impairment and Stealth
The attackers deleted Windows event logs, removed or modified traces containing their IP addresses, disabled shell-history recording, obfuscated commands, altered file timestamps, and disguised malicious components as legitimate operating-system or application files.
Phase 10: Post-Incident Phishing Activity
After defenders began isolating the compromised environment, BO Team attempted to target employees with phishing messages concerning disconnected mail accounts. Investigators also identified abuse of a compromised DKIM signing key, demonstrating that the attackers continued seeking alternative access after containment began.
Indicators of Compromise (IOCs)
| Indicator Type | Indicator | Context |
|---|---|---|
| IP address | 193.3.23[.]236 |
Infrastructure associated with the investigated activity. |
| IP address | 80.76.42[.]139 |
Infrastructure associated with the investigated activity. |
| IP address | 95.81.115[.]77 |
Infrastructure observed during the intrusion. |
| Domain | ip-223-6.dataclub[.]info |
Infrastructure connected with post-incident phishing activity. |
| Domain | lo1yaizaeghaevarooD1ahtaichootae.nofrillsdns[.]com |
Attacker-controlled infrastructure. |
| Domain | lo1yaizaeghaevarood1ahtaichootaz.nofrillsdns[.]com |
Attacker-controlled infrastructure. |
| Domain | vmware-timecheck-api.nofrillsdns[.]com |
Infrastructure associated with disguised VMware-related activity. |
| File path | /etc/cron.hourly/pgcache |
Backdoor disguised as a PostgreSQL-related component. |
| File path | /etc/cron.hourly/1vmware-timesync |
Suspicious persistence artifact. |
| File path | /usr/bin/vmware-toolbox-setdate |
Suspicious file imitating a VMware utility. |
| File path | /dev/shm/pulse-shm-3088612174 |
Shared-memory artifact used to store intercepted credentials. |
| MD5 | 346b04985ec6c5d107ba0ddf32fe7691 |
Linux backdoor associated with the compromise. |
| MD5 | b86690213f0155f7b3b65832301ed932 |
Backdoor associated with the second virtualization cluster. |
| MD5 | 5aa8c19c897176502cf8d8fc543fafc7 |
Obfuscated shell component associated with payload execution. |
| MD5 | 68957de915e2a58339fa477f8078bdf4 |
File associated with the investigated compromise. |
| MD5 | cb2908ef4cc042bfe856c84e9b61a2f9 |
File associated with the investigated compromise. |
| MD5 | 719e813e527920e83349e061b5b1590f |
File associated with the investigated compromise. |
| MD5 | 86ff08749e77d5d078e929e5f9748504 |
File associated with the investigated compromise. |
| MD5 | 5e04b2891635cbaa83ad619fe248e85b |
File associated with the investigated compromise. |
| MD5 | 6d12efa1db15121f88d58b9952e301ef |
File associated with the investigated compromise. |
Defenders should correlate these indicators with authentication records, network telemetry, file-creation events, cron modifications, hypervisor activity, and other behavioral evidence. An indicator match alone does not confirm compromise.
MITRE ATT&CK Coverage
| Tactic | Technique | Technique ID | Observed Behavior |
|---|---|---|---|
| Reconnaissance | N/A | N/A | No supported pre-compromise reconnaissance technique was documented. Internal network scanning occurred after BO Team had already obtained access and is mapped under Discovery. |
| Resource Development | N/A | N/A | No sufficiently supported Resource Development technique was established. |
| Initial Access | Valid Accounts | T1078 | A service account was used to authenticate to an internet-accessible Linux terminal server over SSH. |
| Execution | Command and Scripting Interpreter: Unix Shell | T1059.004 | Shell commands and scripts were executed across Linux and virtualization systems. |
| Persistence | Scheduled Task/Job: Cron | T1053.003 |
Backdoors were stored under /etc/cron.hourly/ for recurring execution.
|
| Privilege Escalation | N/A | N/A | Root-level execution was associated with likely Log4Shell exploitation, but the evidence does not establish a separate transition from lower to higher privileges on the affected system. |
| Defense Impairment | Disable or Modify Tools: Clear Windows Event Logs | T1685.005 | Windows event logs were cleared or overwritten to obstruct investigation. |
| Credential Access | OS Credential Dumping: LSASS Memory | T1003.001 | BO Team attempted to obtain credentials from LSASS memory and subsequently used operating-system components to generate memory dumps. |
| Discovery | Network Service Discovery | T1046 | After obtaining access, BO Team scanned internal address ranges for SMB, Kerberos, SSH, Telnet, FTP, HTTP, HTTPS, WinRM, LDAP, and other network services. |
| Lateral Movement | Remote Services: SSH | T1021.004 | SSH connections and stolen private keys supported movement between Linux systems, hypervisors, and orchestration infrastructure. |
| Collection | Data from Local System | T1005 | BO Team collected credentials, keys, configuration data, source code, and other information from compromised systems. |
| Command and Control | Proxy | T1090 | Proxying and remote-access utilities supported communication through compromised infrastructure. |
| Exfiltration | N/A | N/A | Confidential information was reportedly stolen, but the available evidence does not establish a specific exfiltration technique. |
| Impact | N/A | N/A | Although BO Team has previously deployed ransomware, this incident did not confirm encryption, destructive activity, or malicious software distribution. |
| Stealth | Timestomp | T1070.006 | File timestamps were modified to make malicious artifacts appear older or resemble legitimate system components. |
Mitigation & Recommended Actions
- Patch Log4Shell exposure: Identify Log4j components throughout virtualization appliances, Java applications, development tools, and bundled runtimes. Apply supported vendor updates instead of assuming that updating a system-wide Java installation remediates embedded components.
- Update VMware infrastructure: Upgrade vCenter Server and ESXi systems to supported, fully patched releases. Confirm that historical Log4Shell remediation and subsequent security updates have been applied.
- Restrict administrative interfaces: Keep vCenter, ESXi, GitLab, Ansible, SSH, code-signing systems, and other management services off the public internet. Limit access through dedicated management networks, VPNs, jump hosts, and allowlisted administrative systems.
- Harden SSH authentication: Disable password authentication where operationally possible, require managed cryptographic keys and multifactor authentication, and review service accounts for unnecessary interactive access.
- Protect development and signing secrets: Rotate SSH keys, API tokens, deployment credentials, DKIM keys, certificates, code-signing keys, and automation secrets exposed to compromised systems. Do not assume that password rotation alone removes persistent access.
- Audit code-signing infrastructure: Review build histories, signing logs, certificate usage, generated binaries, security detections, and source-code changes for unauthorized activity. Revoke and replace signing material if exposure cannot be ruled out.
- Inspect virtualization systems: Review ESXi and vCenter authentication, shell, process, file-creation, cron, networking, and virtual-machine management activity for unauthorized changes.
- Hunt for persistence: Search for unexpected files under cron directories, suspicious VMware-themed filenames, modified timestamps, unauthorized SSH keys, unfamiliar virtual interfaces, and remote-access utilities.
-
Investigate credential theft:
Review LSASS access, memory-dump creation, registry hive access, retrieval of
ntds.dit, Kerberos ticket activity, and unusual administrative authentication across domains. - Rebuild confirmed compromises: Rebuild affected management appliances and critical servers from trusted media when root-level compromise is established. Rotate credentials and secrets before reconnecting rebuilt systems.
- Prepare for follow-on phishing: Monitor for messages referencing disconnected mail accounts, validate DKIM signing infrastructure, and revoke compromised mail-signing keys immediately.
Instantly Fix Risks in Endpoints & Cloud with Saner
Saner is a continuous and integrated vulnerability and exposure management platform that identifies vulnerable assets, prioritizes exploitable risks, and helps organizations deploy and verify security updates across endpoints and cloud infrastructure.
Schedule a Demo



