Check Point Software has shipped emergency hotfixes for CVE-2026-123039, a high-severity local privilege escalation flaw in the Gaia OS DynamicCLI IPC interface that attackers have been exploiting since late October to gain root and establish persistence. The vendor says the active campaign has touched a couple of dozen customers, and in at least one intrusion the bug served as the pivot in a chain running from the Mobile Access portal to a destructive wiper on a VMware ESXi host. Firewall, Security Gateway, Security Management Server, Multi-Domain Security Management Server, and Scalable Platforms are all affected.

The flaw in DynamicCLI

The vulnerable component is the DynamicCLI IPC path that carries privileged command requests to the Gaia OS DynamicCLI daemon, dynamic_server. That daemon runs with root privileges and listens on a local socket at /tmp/dynamic_socket.

Two design choices turn that into a privilege escalation primitive:

  • The socket is world-writable, so any local process can connect to it.
  • The daemon does not authenticate the processes that connect.

As a result, an attacker who already holds an authenticated local account can write a command to the socket and have it executed as root — no additional authorization step, no separation between the caller's privileges and the daemon's. Check Point rates the issue high severity and disclosed it in a Thursday security advisory alongside confirmation of in-the-wild use.

Affected products

The vendor lists the following as vulnerable to exploitation of CVE-2026-123039:

  • Check Point Firewall
  • Multi-Domain Security Management Server
  • Scalable Platforms
  • Security Gateway
  • Security Management Server

Attack chain: portal access to ESXi wiper

Check Point's separate incident report describes how the zero-day was used in a broader intrusion. The sequence, as documented by the vendor:

  1. Initial access — attackers reached the Mobile Access portal by exploiting the PHP2Shell vulnerabilities.
  2. Privilege escalation — from that foothold they abused CVE-2026-123039 to move to root.
  3. Persistence — webshells were dropped on the compromised systems.
  4. Tooling — the operators installed the TSGA backdoor and TSProxy malware, both previously observed during investigations into attacks on Ivanti appliances.
  5. Lateral movement — the intruders pivoted from the compromised gateway to a VMware ESXi host.
  6. Impact — destructive wiper malware was deployed on the ESXi host.

The ESXi wiper stage is covered in more depth in BleepingComputer's write-up of the PHP2Shell campaign, and the TSGA/TSProxy tooling overlaps with the activity described in reporting on the Ivanti Connect Secure zero-days.

Detecting exploitation

Check Point has published indicators of compromise and recommends that defenders hunt for three classes of evidence:

  • Commands associated with exploitation attempts. The advisory points to a set of command patterns linked to known attack traffic; matching those against process accounting, audit logs, or command history on Gaia systems is the most direct check.
  • Use of the system() sink. Payloads that route through system() are uncommon — often entirely absent — in legitimate traffic, so any occurrence on a gateway or management server deserves scrutiny.
  • Commands that bypass the default Clish wrapper. Normal administrative activity on Gaia goes through Clish. A command that reaches the daemon without that wrapper suggests it was pushed straight into the socket and executed as root, which is exactly the behavior this flaw enables. Treat such commands as requiring further investigation rather than as routine admin work.

Mitigation

Applying the CVE-2026-123039 hotfixes is the primary fix and should be treated as urgent given confirmed exploitation. For environments that cannot patch immediately, Check Point's interim guidance is to stop the dynamic_server process and prevent it from restarting, which removes the exposed root-execution path at the cost of DynamicCLI functionality. Because that step is a stopgap rather than a fix, the process will come back unless its restart is actively blocked, and the hotfix should follow as soon as the maintenance window allows.

Technical background: why a world-writable root socket is dangerous

The pattern behind CVE-2026-123039 is a common one in appliance and network-OS code: a privileged daemon exposes a Unix domain socket for local clients, and the socket's filesystem permissions become the only access control. If the socket mode allows writes by everyone and the daemon trusts whatever arrives on it, then the daemon's privilege level effectively belongs to any local user.

On Linux, auditing for this class of exposure starts with the socket's metadata and the process holding it:


ls -l /tmp/dynamic_socket
stat -c '%A %U:%G %n' /tmp/dynamic_socket
ss -xlp | grep -F dynamic_socket

A hardened design does not rely on the file mode alone. The socket is typically restricted to a dedicated group (0660 root:<authorized-group> rather than world-writable), and the daemon verifies the connecting peer's credentials — for example with SO_PEERCRED — before honoring a request. Where the protocol permits it, the daemon should also drop privileges after startup and validate or allow-list the commands it will run, so that a single missing check does not translate directly into root code execution. These are general hardening principles, not a description of Check Point's specific implementation.

Prior Check Point zero-days and CISA's catalog

This is not Check Point's first zero-day of the year. In September the company warned that attackers were exploiting two critical remote code execution flaws — CVE-2026-93616 in Security Management Server and CVE-2026-85102 in Security Gateway — in attacks that required no privileges on the target. Coverage of those advisories is available in BleepingComputer's report on the Management Server zero-day and its report on the Security Gateway VPN RCE flaw.

Over the past two years, CISA has added five Check Point vulnerabilities to its Known Exploited Vulnerabilities catalog, and two of those have also been used by ransomware operators. The current catalog entries for the vendor can be filtered in the CISA KEV catalog.

References