A 2021 Log4Shell Bug in VMware vCenter Handed BO Team Every Hypervisor at a Russian Software Maker, Then Its Code-Signing Server Came Under Attack. Check Your vCenter Build Today
If you run VMware vCenter, look up its build number today. On the 7.0 line, anything older than 7.0 Update 3c (build 19234570) still carries Log4Shell, CVE-2021-44228, a bug that has been in CISA's Known Exploited Vulnerabilities catalog since Friday, December 10, 2021. On 6.7 the fix is Update 3q, and on 6.5 it is Update 3s. If you find a vulnerable build, assume anyone who could reach that vCenter could have taken root on it, and on every hypervisor it manages. Then move off vSphere 7 entirely: Broadcom ended general support for it on October 2, 2025, so it no longer gets security patches.
That is the lesson of a report Kaspersky's ICS CERT team published on Thursday, November 5, 2026, and which The Hacker News picked up this morning. A Russian software developer called Kaspersky in June 2025 about strange SMB logins. Kaspersky found that a group it attributes to BO Team, also known as Black Owl, had taken nearly the whole company: domains, hypervisors, mail, VPN, backups, the SIEM, the GitLab server and the server that signs the company's software builds. Kaspersky's reading is that the goal was a supply-chain attack on the developer's customers, which it says are mostly industrial companies in Eastern Europe. It says the attack was stopped before that happened.
What Kaspersky confirmed
The way in was not Log4Shell. The first event Kaspersky could attribute to the intrusion was a login to the developer's Linux terminal server over SSH, which was reachable from the internet, with a service account that accepted a username and password. From there the attackers scanned the internal network, tried and failed to log in to the vCenter web interface, and looked up vCenter documentation and a vulnerability database.
Shortly afterward, root processes appeared on the vCenter appliance with no user session and no SSH login behind them. Kaspersky found that the server ran vCenter 7.0.3, that a vulnerable log4j library was loaded at the time, and that the command matched a Log4Shell demonstration posted on X with only the target address changed. It built a test bench on the same vCenter build and the exploit worked. Kaspersky's own words are that "all of this points to likely exploitation" of CVE-2021-44228. That is a strong inference backed by a reproduction, not a captured exploit request.
With root on vCenter, the attackers read the private key at /root/.ssh/id_rsa and used it to log in to the hypervisors. They planted an hourly cron backdoor at /etc/cron.hourly/pgcache, disguised as a PostgreSQL component and backdated to look like it was last changed in 2021, then used the hypervisor's certificate to reach a second virtualization cluster and plant another. Inside the virtual machines they took the GitLab server, and with its key the Ansible orchestration server, and with the Ansible key the VPN, backup, network-configuration backup and SIEM servers.
On the Windows side they found domain passwords in service scripts any user could read, and a proxy account password sitting in a public LDAP comment field. They dumped lsass memory with a built-in Windows DLL after Nanodump was blocked, copied the domain database, and logged in as a domain administrator. One administrator used the same password in two domains, which gave them the root of the forest.
The code-signing server
Kaspersky says the server that signs the developer's compiled builds drew particular interest, and names two reasons an attacker wants one: to slip code into a build just before it is signed, or to steal the signing key and sign malware with it later. The security product's logs on that server had been deleted, but Kaspersky recovered three blocked detections showing attempts to run a WMI script, create a scheduled task and drop malware. Kaspersky does not say the signing key was stolen, and does not say any signed build was altered. Press headlines that the server was "compromised" go further than the report's text on what was taken.
A different key was taken. After the developer cut its network off from the internet, employees received phishing mail titled "Account disconnected from the mail server". It passed SPF and carried a valid DKIM signature, which Kaspersky reads as proof the attackers held the private key from the company's own mail server. Kaspersky says the phishing was caught before credentials leaked.
What a stolen signing key would mean
If you sign software, the point of the key is that customers' machines trust what it signs. Anything signed with a stolen key verifies as genuine, and nothing in the signature tells a customer which builds you made and which the attacker made. Revoking the certificate stops new trust, but how older signatures are treated after revocation depends on the platform and on timestamping, so the cleanup is long and public. The defenses that would have held here: keep the key in a hardware security module so it cannot be copied off disk, keep the signing server off the Windows domain the rest of the company uses, and log every signature so you can reconcile each one against a build your pipeline actually produced. If you buy software from a vendor, a valid signature tells you who signed it, not that the vendor's build system was clean.
What is not known
The report does not name the developer or its product, and does not say whether any build reached customers altered. It does not give a date for the first intrusion. The developer called Kaspersky in June 2025, the phishing wave came at the end of July, and Kaspersky says it waited almost a year to publish so the developer could fix its weaknesses. Kaspersky attributes the attack to BO Team with "a high degree of confidence", from a credential-dumping command matching an earlier BO Team intrusion character for character, its obfuscation and persistence habits, and VPN connections from two addresses in a Lviv, Ukraine provider's pool. That is Kaspersky's assessment, and we have no independent evidence for or against it.
What to do
Find every vCenter and read its build. The vSphere Client's About box and the appliance management page on port 5480 both show it. On 7.0, anything below Update 3c, build 19234570, is vulnerable to CVE-2021-44228. If you cannot patch at once, VMware's KB 87081 has the interim workaround, but on vSphere 7 the real fix is moving to a supported release.
Do not let user networks reach management interfaces. vCenter, ESXi and the appliance pages belong on their own segment, reachable only from a short list of admin hosts. Here the attackers reached vCenter from an ordinary terminal server.
If a vCenter ran a vulnerable build, treat its keys as stolen. Rotate the root SSH keys on vCenter and the hosts, check /etc/cron.hourly and other cron directories on hypervisors and Linux VMs for files you did not put there, and look for the file paths in Kaspersky's list below.
Close password SSH from the internet. Keys only, root login off, a limit on attempts, and VPN in front. The attackers also guessed a root password after about 2,500 tries on an internal server.
Hunt for stored passwords. Scripts on shares, LDAP comment and description fields, text files. After a domain compromise, reset the krbtgt account password twice.
Protect your signing key and your mail key. Hardware security module for code signing, signer off the domain, every signature logged. Rotate DKIM selectors after any mail-server compromise.
Block the indicators below. They are now in our feed.
How we know
We read Kaspersky ICS CERT's report in full on November 16, VMware KB 87081 for the fixed vCenter builds, Broadcom's notice on the end of general support for vSphere 7.0, and the CISA KEV JSON feed (catalog release of November 13, 2026) for the Log4Shell entry: added December 10, 2021, due December 24, 2021, known ransomware use. The Hacker News article was our lead, published November 16 at 06:53 UTC; every fact here comes from Kaspersky's report. The Orange Cyberdefense profile of BO Team that The Hacker News cites is background we did not rely on.
We checked our own edge for Log4Shell lookup strings and vCenter login paths. Across 444,104 edge-block records, 75,829 edge-canary records and 1,855 origin-honeypot records since October 1, none matched. That is a narrow measurement. We store the path and user agent of blocked requests, not every header, and Log4Shell strings often ride in other headers, so we could miss them. We run no vCenter decoy. And in this intrusion vCenter was attacked from inside the network, which no internet sensor sees.
In our feed
We added Kaspersky's published indicators, cited to Kaspersky and not as our own observations, under the source manual-batch-bo-team-log4shell-vcenter-code-signing:
Nine MD5 file hashes, confidence 85: the cron backdoors and downloader scripts, the ReverseSSH tool built from a public GitHub project with the attackers' key, and the Global Socket downloader. 346b04985ec6c5d107ba0ddf32fe7691, 5aa8c19c897176502cf8d8fc543fafc7, 5e04b2891635cbaa83ad619fe248e85b, 68957de915e2a58339fa477f8078bdf4, 6d12efa1db15121f88d58b9952e301ef, 719e813e527920e83349e061b5b1590f, 86ff08749e77d5d078e929e5f9748504, b86690213f0155f7b3b65832301ed932, cb2908ef4cc042bfe856c84e9b61a2f9.
Three command-and-control domains, confidence 85, each a subdomain of the free dynamic DNS service nofrillsdns.com. We list the exact subdomains, not the service: lo1yaizaeghaevarood1ahtaichootae.nofrillsdns.com (Kaspersky prints it with a capital D; DNS ignores case, so we list it lowercase), lo1yaizaeghaevarood1ahtaichootaz.nofrillsdns.com and vmware-timecheck-api.nofrillsdns.com. The attackers parked them at 127.0.0.1 when idle, so a lookup that returns localhost is not proof they are dead.
Three IP addresses, confidence 80: 95.81.115.77, the command server the domains pointed to after detection, and 193.3.23.236 and 80.76.42.139, the attackers' VPN sources. These were seen in 2025 and may since have been reassigned.
We left out ip-223-6.dataclub.info, the hosting-provider server that sent the phishing, because it is a rented host under a provider's domain and was used once in 2025. Kaspersky also lists file paths: /etc/cron.hourly/pgcache, /etc/cron.hourly/1vmware-timesync, /usr/bin/vmware-toolbox-setdate and /dev/shm/pulse-shm-3088612174, the last being where a tampered /usr/bin/nxnode-login wrote stolen passwords. Search your Linux hosts for them.
Sources
Kaspersky ICS CERT: Supply chain attack: BO Team in a software developer's infrastructure (November 5, 2026)
VMware / Broadcom: KB 87081, workaround and fixed builds for CVE-2021-44228 in vCenter Server
Broadcom: End of General Support for vSphere 7.0
CISA: Known Exploited Vulnerabilities catalog, CVE-2021-44228
The Hacker News: BO Team Exploits Log4Shell in VMware vCenter, Compromises Russian Firm's Code-Signing Server (lead only)
If this was useful, or if it wasn't, the rating widget below takes ten seconds and is the only way we know.
Every indicator in this post is in the feed. Free.
1.58M+ IOCs, STIX 2.1 / TAXII, 88% novel vs ThreatFox, exploited-CVE leads ahead of CISA. No credit card — a free API key in 30 seconds, and you can audit every claim above against the live endpoints.



Comments