CVE-2026-123039 Check Point 0-Day Vulnerability

Root access on Gaia OS, affected versions, fixed Takes and a checklist to confirm your gateway wasn’t compromised.

CVE-2026-123039 is a flaw in Check Point’s Gaia OS that lets someone with local access to a gateway or management server run commands as root. Attackers are already using it in the wild, paired with the PHP2Shell bugs to get that first foothold. Fixed Jumbo Hotfix Takes are out, and there’s a temporary workaround if you can’t patch today.

At a glance

  • What it is: a local privilege escalation bug in the DynamicCLI service of Gaia OS, rated CVSS 7.8
  • Status: exploited in the wild, and on CISA’s Known Exploited Vulnerabilities list since November 13, 2026
  • Who’s affected: Force Firewalls, Management and Log Servers, Multi-Domain systems and Maestro Orchestrator running R81 through R82.20
  • The fix: the November Jumbo Hotfix Takes (table below)
  • Can’t patch yet: stop the dynamic_server process as a stopgap
  • Patched doesn’t mean clean: check for signs the box was hit earlier

If you run Check Point gateways, you probably saw the alert on November 12, then again when CISA added the flaw the next day. Most coverage stops at “patch now.” Fair enough, but it skips the question admins actually lose sleep over: was my gateway already compromised? This guide covers what the flaw is, which versions need what, how to fix it, and how to check.

What is CVE-2026-123039?

Inside Gaia OS, a process called dynamic_server runs as root and handles extended command-line functions. It takes requests through a local socket at /tmp/dynamic_socket. On vulnerable systems that socket is world-writable (permissions 0777), and the service never checks who is sending the request. Whatever arrives gets executed as root.

In plain terms, any local user, or any process an attacker controls, can hand the box a command and get root back. Check Point scored it 7.8 on the CVSS scale. That reads as moderate, but a score for a local flaw says little about how it behaves inside a chain.

One thing to get straight early: this is not a remote exploit on its own. The attacker needs code execution on the appliance first. That makes the Check Point Gaia DynamicCLI vulnerability a second-stage tool, which is exactly why the first stage matters so much.

How attackers used it: the PHP2Shell attack chain

Check Point Research published a full write-up of the campaign, including indicators of compromise. The sequence looked like this:

  1. Entry. Attackers exploited PHP2Shell (CVE-2026-109618 and CVE-2026-109619, two critical PHP flaws) against internet-facing Mobile Access portals. That gave them code execution as the portal’s service account, not as root.
  2. Escalation. They used CVE-2026-123039 to jump from that limited account to root.
  3. Staying power. Web shells went into a directory the portal uses to build its login pages. They also installed a backdoor called TSGA, disguised as a Check Point process named CPconfigdaemon, plus a proxy tool called TSProxy.
  4. Damage. In the first wave, they moved on to a VMware ESXi host and ran a wiper, then left a note using the Dream Security name.

Timeline

  • October 21: PHP patches PHP2Shell
  • October 26: PHP2Shell lands on CISA’s KEV list
  • October 30: first wave begins, aimed at Israeli organizations, with destructive activity at some of them
  • November 2: a broader wave hits organizations in several countries, with no destruction observed
  • November 12: Check Point discloses CVE-2026-123039 and ships fixes
  • November 13: CISA adds it to the KEV catalog

Who was behind it is still unsettled. Check Point says the evidence doesn’t support a high-confidence call. The activity ran during typical Chinese working hours and used China-linked tooling, yet the wiper doesn’t fit known Chinese operations. The actor could be an access broker, a false flag, or the destruction could be a smokescreen. If a headline names a country as fact, be skeptical.

Is the Check Point root access vulnerability exploited right now?

Yes. Check Point says it saw exploitation against roughly two dozen customers. CISA added the flaw to its Known Exploited Vulnerabilities catalog on November 13 and set a November 16 due date for US federal civilian agencies. That date has already passed. The deadline only binds federal agencies, but it’s a decent signal of how urgent this is for everyone else.

Which Check Point versions are affected, and which Takes fix it?

Affected products are Force Firewalls (Security Gateways), Management and Log Servers, Multi-Domain management and logging systems, and Maestro Orchestrator across R81, R81.10, R81.20, R82, R82.10 and R82.20. Spark Firewalls and Smart-1 Cloud are not affected.

In plain terms, any local user, or any process an attacker controls, can hand the box a command and get root back. Check Point scored it 7.8 on the CVSS scale. That reads as moderate, but a score for a local flaw says little about how it behaves inside a chain.

One thing to get straight early: this is not a remote exploit on its own. The attacker needs code execution on the appliance first. That makes the Check Point Gaia DynamicCLI vulnerability a second-stage tool, which is exactly why the first stage matters so much.

Release Fixed from Notes
R82.20
Jumbo Hotfix Take 14
R82.10
Jumbo Hotfix Take 52
R82
Jumbo Hotfix Take 132
R81.20
Jumbo Hotfix Take 174
R81.10 and R81
No fix
End of support. Use the workaround and plan an upgrade.

Vendor and government notices don’t all list identical ranges, so confirm your exact role and release in Check Point’s advisory sk1000362 before you schedule the work.

How to patch CVE-2026-123039 (and what to do if you can’t yet)

  • Install the Jumbo Hotfix at or above the Take listed for your release. Start with gateways that expose Mobile Access to the internet, since that’s where the attacks began.
  • Don’t stop at the earlier PHP2Shell hotfix. It closed the front door, but this flaw needs the November Take or the workaround.
  • Verify afterward. Once patched, /tmp/dynamic_socket should show 0660 permissions, owned by root and the config group, instead of 0777.
  • If you can’t patch today, run dbset process:dynamic_server, then dbset :save. This stops the service and keeps it from restarting. The catch is that Check Point’s documented Dynamic CLI commands stop working too, so check that nothing in your automation depends on them. Treat it as a bridge, not a fix.
  • On R81 or R81.10, apply the workaround now and put the upgrade on the calendar.

How to check if your Check Point gateway is compromised

A patch closes the door. It doesn’t remove anything that already got inside. Between Check Point’s guidance and its research report, this is a sensible order of work:

  1. Read /var/log/dynamicCli.log. The daemon records the commands it receives. Requests missing the expected Clish wrapper deserve a closer look.
  2. Hunt for the usual persistence moves: new accounts, added SSH keys, unexpected downloads, access to sensitive files.
  3. Inspect the Mobile Access portal directories for web shells, meaning files that have no business sitting in the folder used to build login pages.
  4. Look for process impersonation, especially a backdoor posing as CPconfigdaemon.
  5. Review firewall logs for gaps. Log tampering was part of this toolset, so a missing stretch of logs is a finding, not something to shrug off.
  6. Check ESXi hosts the gateway can reach for unexpected SSH logins or signs of wiping.
  7. Compare against the published indicators of compromise in the Check Point Research report.

If the logs you need don’t exist, mark the gateway “unresolved,” not “clean.” The TSGA implant is reported to steal configuration data, so rotate secrets stored on or reachable from any gateway you can’t clear.

What this incident teaches beyond Check Point

  • The window was short. The PHP bugs were patched on October 21, and attacks on Check Point gateways began nine days later.
  • Edge devices are where chains start. A firewall holds configs, VPN sessions and routes. Root on it is a launchpad, not just a foothold.
  • Segment hypervisor management. The pivot from gateway to ESXi is what turned an intrusion into data destruction. Limit which hosts can reach ESXi management at all.
  • Shrink what a first-stage account can do. Tight identity and privileged access controls limit how far a compromised service account can travel, which is the whole game when the second stage is a privilege escalation.
  • This isn’t Check Point’s first this year. A VPN authentication bypass in June and a SmartConsole zero-day in July were also exploited. Plan for edge-device patching as a routine job, not an emergency.

Need a second opinion on your exposure?

If you’re not sure whether your gateways were exposed, or you want a second pair of eyes on how much damage a compromised appliance could do in your environment, talk to the Hassium team. Bring your Gaia release and Take numbers and we’ll start from there.

Patched your Check Point gateway, but still not sure it’s clean?

A patch closes the door, but it doesn’t remove a backdoor that’s already inside. Hassium helps organisations tighten access controls, protect privileged accounts and reduce the risk of credential-based attacks.

Frequently Asked Questions (FAQs)​

What is CVE-2026-123039?

A local privilege escalation flaw in the DynamicCLI service of Check Point Gaia OS. A world-writable, unauthenticated socket lets a local user run commands as root. It scores 7.8 on CVSS and is exploited in the wild.

Not by itself. The attacker needs local code execution first. In the observed attacks, that came from PHP2Shell on Mobile Access portals.

R82.20 Take 14, R82.10 Take 52, R82 Take 132 and R81.20 Take 174 or later. R81 and R81.10 are end of support and get no fix.

Yes, temporarily. Run dbset process:dynamic_server and dbset :save to stop the service. This disables the Dynamic CLI commands Check Point documents, so test the impact first.

No. The earlier hotfix addresses the entry point. CVE-2026-123039 needs the November Jumbo Hotfix Take or the workaround.

Review /var/log/dynamicCli.log, look for new accounts, SSH keys and web shells, check for a fake CPconfigdaemon, and compare against the indicators in Check Point Research’s report.