Supply Chain · Vulnerability
BO Team Exploited Log4Shell in VMware vCenter to Reach a Code-Signing Server
CVE-2021-44228, Log4Shell, was disclosed in December 2021 and patched within days of disclosure. By June 2025, a threat actor called BO Team walked through that exact vulnerability in an unpatched VMware vCenter Server 7.0.3 instance, pivoted to the underlying ESXi hypervisor over SSH, installed a persistent hourly cron backdoor, and ultimately reached the victim's CI/CD servers and code-signing infrastructure. Kaspersky ICS CERT published its full report on 5 November 2026, nearly eighteen months after the initial engagement with the affected Russian software developer. The gap between patch availability and exploitation in this case was more than three years. That is the detail that should give every infrastructure team pause.
What actually happened
The affected organization was a Russian software developer whose identity has not been publicly disclosed. According to Kaspersky ICS CERT, the company contacted Kaspersky in June 2025 after detecting suspicious activity. Investigators attributed the intrusion to BO Team, a group also tracked under the aliases Black Owl, Hoody Hyena, and Lifting Zmiy. The group is assessed to be of Ukrainian origin, active since at least 2023, and focused on espionage and hacktivist operations against Russian entities. BO Team publicly debuted on a dedicated Telegram channel in early 2024 and has been linked by Kaspersky to coordination with another hacktivist group, Head Mare, including shared infrastructure, as confirmed in a May 2026 report.
The entry point was CVE-2021-44228, the Log4Shell remote code execution vulnerability in Apache Log4j, affecting VMware vCenter Server versions 6.5.x, 6.7.x, and 7.0.x. The victim was running vCenter 7.0.3 without the available patch. Exploiting that vulnerability gave the attackers a remote command line with root privileges on the vCenter host. From there, they connected to the underlying ESXi hypervisor over SSH, a natural pivot given that a compromised vCenter with root access to the management plane is often positioned to reach the hypervisor directly.
On the ESXi host, BO Team planted a shell script named pgcache in the /etc/cron.hourly/ directory. The name appears chosen to blend in with legitimate PostgreSQL-related process names. Kaspersky detected the script as HEUR:Backdoor. The script ran every hour and checked for an activation signal delivered through a DNS record change, allowing the attackers to trigger malware downloads remotely without maintaining a persistent outbound connection that might trip detection rules. The compromise eventually extended to the victim's CI/CD servers and software code-signing server, raising the possibility of a downstream supply chain attack. No malicious software release was confirmed as a result of the compromise, though whether any code-signing certificates or private keys were exfiltrated could not be confirmed from available reporting.
BO Team is known to deploy ransomware, publicize operations on Telegram, and engage in what researchers describe as cognitive warfare aimed at humiliating Russian institutions. Its toolkit includes Mythic, Cobalt Strike, and its own malware families BrockenDoor and ZeronetKit.
Why this should matter to you even if you don’t run VMware vCenter
The instinct when reading this story is to treat it as geopolitically specific: a Ukrainian hacktivist group attacking a Russian company. That framing misses the operational lesson. The entry point was a three-year-old vulnerability in VMware vCenter that had a vendor patch available almost from the moment it was disclosed. The geopolitics are incidental to the technique. Any organization running unpatched vCenter is exposed to the same initial access path, regardless of who is on the other side.
VMware vCenter is one of the most common hypervisor management platforms in enterprise environments, and it sits at a privileged position in the infrastructure stack. Compromise of vCenter frequently means access to every virtual machine it manages, plus the underlying hypervisors. Attackers have known this for years, which is why vCenter has been a repeated target across ransomware campaigns, nation-state intrusions, and now hacktivist operations. The pattern in this case, exploiting a known CVE to get root on the management plane, then pivoting laterally to reach build and signing infrastructure, is not unusual. It is a well-worn path.
The code-signing server detail is worth dwelling on. A signing server that is reachable from the same network segment as a compromised hypervisor, without additional segmentation or access controls, represents a concentration of risk. Signed software carries an implicit trust signal to downstream users and security tools alike. An attacker who can influence what gets signed, or steal the signing keys, can potentially distribute malicious code that passes automated verification. Even without a confirmed malicious release in this case, the exposure was real and the blast radius would have extended well beyond the victim organization.
Organizations running VMware vCenter should also factor in the ESXi pivot that followed the initial compromise. Attackers moved from vCenter to ESXi over SSH, which is a path that requires SSH to be enabled and accessible on the hypervisor host. Many environments leave ESXi SSH enabled for operational convenience. That is a tradeoff worth revisiting, particularly for systems that are not under active maintenance.
The control that would have blunted it
The primary control failure here was patch management. CVE-2021-44228 had a patch available, and the affected system was still running a vulnerable version more than three years later. A disciplined patch management program, one with defined SLAs for critical and actively exploited vulnerabilities, would have closed this window. For context, Log4Shell carried a CVSS score of 10.0 and was added to CISA's Known Exploited Vulnerabilities catalog in December 2021. For vulnerabilities at that severity level, most mature programs target remediation within days to weeks. Cyber insurers increasingly ask for documented evidence of patch cadence and prioritization processes, particularly for CVEs on the KEV catalog. If you cannot show that evidence, expect questions at renewal.
Patch management on vCenter specifically carries some operational weight. VMware vCenter updates can be disruptive to production workloads and require maintenance windows, compatibility testing against guest VMs, and coordination with application teams. Those are real constraints. They explain how patching slips, though they do not make the resulting exposure acceptable. Compensating controls matter here: vulnerability scanning with prioritized alerting on critical CVEs, network segmentation that limits what an attacker can reach from a compromised vCenter instance, and monitoring for anomalous SSH sessions from the management plane to ESXi hosts are all measures that reduce the blast radius when patching is delayed.
The persistence mechanism in this case, a cron job in /etc/cron.hourly/ named to mimic a legitimate process, is exactly the kind of thing that file integrity monitoring catches. FIM on ESXi hosts, particularly watching cron directories and system binary paths, would likely have surfaced the implant earlier. This is a control that many teams skip on hypervisor hosts because those hosts are treated as appliances rather than managed systems. Treating ESXi as an unmonitored appliance is an assumption worth revisiting.
Network segmentation between the CI/CD environment, the code-signing server, and the broader infrastructure is also directly relevant here. Code-signing infrastructure warrants isolation, privileged access controls, and audit logging as a matter of course. If the signing server cannot be reached from a compromised hypervisor management host without crossing an explicit access boundary, the attacker's path to supply chain impact becomes significantly harder. That segmentation does not require exotic tooling. It requires treating the signing infrastructure as a tier-zero asset with matching controls.
Does Your Cyber Insurance Require Proof of Patch Management?
Download the Strondex Cyber Insurance Readiness Checklist to see exactly what patch cadence evidence, vulnerability scanning records, and compensating control documentation your insurer is likely to ask for.
If you want to map your current patch management and segmentation posture against what cyber insurers actually require, the Strondex Cyber Insurance Readiness Checklist gives you a structured starting point.
Sources
- The Hacker News - "BO Team Exploits Log4Shell in VMware vCenter, Compromises Russian Firm's Code-Signing Server": thehackernews.com
- Broadcom/VMware Knowledge Base - "CVE-2021-44228 vCenter Advisory": knowledge.broadcom.com
- Orange Cyberdefense - "BO TEAM Group Report": orangecyberdefense.com
- INCYBER News - "Ukrainian Hacktivist Groups BO Team and Head Mare Cooperate Against Russian Targets": incyber.org
Reported figures vary by source and were accurate as of publication; this article is general security commentary, not specific security or underwriting advice.