Kaspersky links BO Team breach to software supply-chain threat

Kaspersky links BO Team breach to software supply-chain threat

Add CyberWorldOpsto your preferred sources on Google
Kaspersky links BO Team breach to software supply-chain threat
APT

Illustrative image generated with AI

Attackers took control of a Russian software developer’s development systems and code-signing server in an intrusion that Kaspersky believes was intended to compromise the company’s customers. In a report published November 5, 2026, Kaspersky ICS CERT attributed the operation to BO Team and said responders apparently stopped it before the attackers achieved that objective.

The unnamed developer serves customers in Russia. The case concerns a months-long compromise of one organization, with consequences for software developers protecting build systems and administrators running vulnerable VMware vCenter installations: the attackers exploited Log4Shell during their movement through the network, after first entering through an internet-accessible SSH service.

The developer contacted Kaspersky in June 2025 after administrators’ computers recorded unexpected SMB connections—attempts to access Windows network services—from accounts that should not have reached those machines. Investigators found that attackers had seized nearly all key infrastructure, including the Windows domain, servers, network equipment, email, virtualization, backups, security systems, web servers and DNS administration.

From exposed SSH to development infrastructure

Kaspersky identified the earliest reliably attributable event as a login to a Linux terminal server through SSH, a service for remote command-line access. The attackers used a service account that permitted username-and-password authentication. They then surveyed internal systems for accessible services, including file sharing, authentication, remote administration and web interfaces.

The Hacker News, reporting additional findings from Kaspersky, said attempts to sign in to the VMware vCenter web interface failed. The attackers subsequently exploited CVE-2021-44228, known as Log4Shell, on a vulnerable vCenter Server identified as version 7.0.3, obtaining a command line with root privileges, the highest level of system access. The flaw affects the Apache Log4j logging component shipped with vCenter.

The attackers connected to a hypervisor—the system that runs virtual machines—over SSH and installed a backdoor at /etc/cron.hourly/pgcache. Its hourly execution and PostgreSQL-like name helped it persist while resembling legitimate software. It fetched another script from lo1yaizaeghaevarooD1ahtaichootae.nofrillsdns[.]com, which downloaded and ran Global Socket, a utility used here for remote control.

Kaspersky described a simple activation mechanism: while the command server was inactive, its DNS record pointed to 127.0.0.1, the local machine address. Changing that record caused the backdoor to download and execute its payload. Some variants instead requested a host-specific payload from lo1yaizaeghaevarood1ahtaichootaz.nofrillsdns[.]com, whose next stage remained undetermined.

The attackers also accessed the corporate proxy and used a compromised hypervisor’s certificate to reach a second virtualization cluster and install a similar backdoor. They collected virtual-machine and firewall information, ran commands inside some guest operating systems, and used an obfuscated Bash script to decrypt and execute a Linux binary providing remote access over SSH.

Secrets found on compromised virtual machines opened a route into GitLab, the software development platform. The attackers accessed source-code repositories and obtained the GitLab server’s private key, then used it to enter Ansible, the organization’s system for automating administration.

Stolen credentials widened the breach

The operation relied heavily on legitimate accounts and built-in administration tools. Investigators found credential searches, use of Windows utilities to identify security software and network settings, and changes to security attributes on a user’s Downloads directory.

The attackers reached a domain controller using a system administrator’s account, likely after accessing memory from virtual machines on compromised hypervisors. A security product blocked an attempt to use Nanodump against lsass.exe, the Windows process that handles authentication secrets; the attackers then deleted event logs. They also used the Windows component comsvcs.dll to dump that process and exported registry data with regedit.exe to decrypt ntds.dit, the Active Directory database, extracting credentials for all domain users.

Collected information was transferred to the terminal server with OpenSSH’s scp.exe. A domain administrator’s Kerberos ticket—an authentication credential—provided another route to the domain controller. Reuse of the same system-administrator password across different domains ultimately helped the attackers compromise the Windows infrastructure.

Other persistence efforts included modifying /usr/bin/nxnode-login to capture credentials in /dev/shm/pulse-shm-3088612174. Investigators also found root logins to the terminal server, possibly using credentials stolen from an administrator’s computer, and a successful password-guessing attack against an administrator’s terminal server after roughly 2,500 failed attempts, used to place the Ansible system’s private key.

The attackers concealed activity through code obfuscation, altered file timestamps, repeated event-log deletion and malware named to resemble legitimate software. They generally favored existing tools such as RDP, SSH and SCP over conspicuous hacking utilities.

Why the signing server mattered

Kaspersky found traces of access to multiple Windows servers but malware use on only one: the server that signed compiled commercial software builds. Code signing gives customers a way to check a publisher’s identity and whether software has changed since signing. Access to that server could therefore support either inserting malicious code before a build was signed or stealing signing credentials for later malware use.

The researchers suspected an attempt to execute a malicious WMI script, using Windows’ management interface, and install a backdoor in a temporary directory. Their assessment of a possible supply-chain objective also drew on the attackers’ interest in source repositories and use of a software composition analysis platform, which examines software dependencies.

BO Team, also called Black Owl, Hoody Hyena and Lifting Zmiy, has operated since at least 2023 and is assessed to be of Ukrainian origin, The Hacker News reported. The outlet cited Orange Cyberdefense’s assessment that the group seeks to weaken Russia’s military-industrial capabilities and embarrass Russian institutions. Kaspersky observed connections from 193.3.23[.]236 and 80.76.42[.]139, belonging to a Ukrainian provider, and activity following a regular work schedule in a time zone close to the victim’s.

Although BO Team has previously stolen information, deployed ransomware and publicized attacks on Telegram, this operation involved no data destruction or public claim. After responders cut off access, the attackers attempted to regain employee credentials through a phishing campaign in late July.

Those messages used the subject “Account disconnected from the mail server” and directed recipients to a password-change page. Kaspersky said the emails passed SPF and DKIM checks, mechanisms intended to validate email origin, indicating possession of a private key obtained through the mail-server compromise. The phishing service copied the organization’s website and overlaid a password form, sending entered credentials to the attacker while displaying an error. The campaign was detected and prevented.

Check vCenter fixes and respond to exposed credentials

Broadcom’s VMware guidance identifies these fixed vCenter Server releases for CVE-2021-44228 and the related CVE-2021-45046:

  • 7.0 Update 3c, build 19234570.
  • 6.7 Update 3q, build 19300125.
  • 6.5 Update 3s, build 19261680.

Administrators should check the precise release and build. Broadcom describes its workarounds as temporary and recommends the unified vc_log4j_mitigator.py script over manual changes. The script stops and restarts vCenter services; --dryrun, also available as -r, checks for vulnerable files without changing them or restarting services. Do not run the mitigator on an already fixed release. Existing workarounds need not be reversed before upgrading.

Third-party vSphere Client plugins can recreate vulnerable files after remediation. Broadcom says affected plugins must be patched, marked incompatible or removed through the vCenter Managed Object Browser; disabling them in the client is insufficient. Its Cloud Gateway workaround requires only the specified vMon, Analytics and class-removal steps. Separately, vCenter Server Appliance 6.0 U3j is vulnerable through Performance Charts and requires its documented manual procedure because the class-removal script was not verified for 6.0; 6.0 GA through U3i are not vulnerable, although some contain unused vulnerable libraries. The article also requires removing vCenter High Availability before applying its procedures.

For organizations finding indicators of this campaign, Kaspersky recommends immediately isolating the compromised subnet or disconnecting enterprise internet access, monitoring suspicious connections, replacing all user and computer passwords across domains, and reissuing private keys and certificates. Its advice includes resetting the krbtgt account twice to counter forged Kerberos “Golden Tickets,” with a warning that this temporarily disrupts Active Directory services and requires service-account reauthentication. The immediate-response instructions exclude users who only received phishing emails they did not open.

Responders should also reboot affected systems to restore potentially damaged security interception components, run full scans and inspect scheduled tasks. Kaspersky’s published indicators include /usr/bin/vmware-toolbox-setdate, /etc/cron.hourly/1vmware-timesync, vmware-timecheck-api.nofrillsdns[.]com, 95.81.115[.]77 and ip-223-6.dataclub[.]info, alongside file hashes and detection names in its report.

For prevention, Kaspersky recommends supported, patched operating systems, hypervisors, firmware and applications; tighter segmentation and access controls; login-attempt limits; restrictions on interactive service-account logins; and two-factor authentication for critical systems and external services. It also calls for removing plaintext secrets, ending potentially compromised external sessions, strengthening email filtering, and central monitoring of security events. Organizations using its products should apply its security-management hardening guidance, keep protection enabled and current, minimize exclusions, and use Kaspersky Security Network where permitted. Administrators investigating a match should treat credentials, signing material and remote-access paths as part of the recovery, alongside the vulnerable server.

Security dossiers

Read next

Sources

Back to home

Latest Cybersecurity News

All cybersecurity news →