← ข่าวทั้งหมด

บทความเผยแพร่เป็นภาษาอังกฤษ

Threat IntelligenceAttack SurfaceVulnerability Management

Attackers Rooted Check Point Gateways Through the Mobile Access Portal, Then Wiped the VMware Hosts Behind Them. Do You Know Which of Your VPN Portals Face the Internet?

Obiguard Research Team·15 พฤศจิกายน 2569·5 min read

Attackers broke into unpatched Check Point Security Gateways through their internet-facing Mobile Access portals, escalated to root with a newly disclosed flaw, CVE-2026-123039, and in some cases moved on to VMware ESXi hosts and ran a disk wiper. Check Point Research published its analysis on 12 November 2026, and BleepingComputer reported it on 13 November. Initial access came through PHP2Shell (CVE-2026-109618 and CVE-2026-109619), the PHP flaws we covered in our PHP2Shell timeline. This article is based on those two reports, and we have not independently verified the findings.

What happened

  • Start date: Check Point says scanning and exploitation began on 30 October 2026, about a week after a fix became available, against unpatched gateways with Mobile Access portals.
  • Two waves: the first, on 30 October, targeted Israeli organisations and included destructive activity. A second, broader wave on 2 November hit organisations in several countries and sectors, with no evidence of destruction. Check Point describes both as opportunistic.
  • Root on the gateway: PHP2Shell gave code execution as the portal's service account. The attackers then used CVE-2026-123039, a local privilege escalation in the gateway's DynamicCLI interface, to become root.
  • Persistence: PHP webshells were dropped into the Mobile Access templates directory as zz<random>.tpl files, and a backdoor named TSGA ran as CPconfigdaemon, relaunching at boot as slot_CPconfigdaemon.
  • From firewall to hypervisor: the attackers ran a network scanner from the gateway, then connected over SSH to ESXi hosts using the vpxuser account and ran a wiper called destroy_v6. It leaves a ransom-style note, DREAM_ANNEXATION.txt, in each directory it touches.

What is not settled

Attribution. Check Point links the backdoor tooling to modified variants of PhiliKit, which ESET has tied to the China-nexus group UNC5221, but says the evidence does not support a confident assessment. It lists three possibilities: one actor ran the whole campaign, an access broker handed over the gateways, or the political messaging is a false flag. It also notes the wiping may have been a smoke screen.

Affected versions. Neither report lists affected or fixed builds. BleepingComputer says Check Point's November Jumbo Hotfix fixes both PHP2Shell and CVE-2026-123039, and that it should go on all Management Servers and Security Gateways, starting with internet-facing Mobile Access portals. Check Point's support articles are the authority on which takes you need.

Why this one is different

We have already seen PHP2Shell used to move from a firewall to an LDAP service account, and a separate Check Point zero-day that left management and log servers open for two months. What is new here is the destination. A remote-access gateway sits in front of the network and is trusted by what is behind it. In this campaign, that trust ran all the way to the hypervisor, where a single wiper run can take down every virtual machine on a host.

It also means the patch is not the end of the job. A gateway patched today may already carry a webshell or a backdoor that survives reboots.

What to check now

  1. List every Mobile Access and remote-access portal you expose. Include gateways at branch offices, acquisitions and disaster-recovery sites. You cannot prioritise a hotfix for a portal nobody remembers.
  2. Apply the November Jumbo Hotfix. Start with gateways whose portal is reachable from the internet, and confirm the take against Check Point's own guidance.
  3. Hunt on gateways, even patched ones. Per the reports, look for session files containing upload_progress_pwn, unexpected .tpl files in the Mobile Access templates directory, CPconfigdaemon processes or binaries that should not be there, a listener on TCP 11072, and gaps in firewall logs.
  4. Check the path to your hypervisors. BleepingComputer calls unauthorised SSH from a gateway to an ESXi host "a clear sign of breach." A firewall should not be able to SSH to vCenter-managed hosts at all, so make that a rule, not just an alert.
  5. Block the reported infrastructure. Check Point lists 130.94.19[.]204 and 130.94.19[.]168, plus file hashes, in its write-up.
  6. Confirm you can restore ESXi workloads offline. If a wiper reaches the hypervisor, backups that the same hosts can delete do not help.

How Obiguard Scout fits

Step 1 is the one most teams get wrong, because remote-access portals are often set up once and then forgotten. Obiguard Scout is our external attack surface management product. Starting from a domain you verify, it maps the hosts underneath it and what each one appears to be running, then flags changes from one scan to the next. For this campaign, that gives you a list of which public hosts present a VPN or remote-access login page, so the hotfix can go to the exposed gateways first, and a new portal that appears next month does not go unnoticed.

Scout works only from what is visible on the internet. It cannot see whether a gateway already has a webshell, read its logs, tell which Jumbo Hotfix take is installed if the device does not reveal it, or check whether a gateway can reach your ESXi hosts. Steps 3 to 6 still need access inside the network. You can run a free Scout check on a domain you own to see what it already exposes, or get in touch to talk through coverage.

Sources

Obiguard ช่วยได้อย่างไร

เปลี่ยนเรื่องนี้ให้เป็นนโยบายที่บังคับใช้ได้จริง ไม่ใช่แค่การรับรู้

Obiguard อยู่หน้าทุกคำขอ AI ขององค์กรคุณ คัดกรองพรอมต์และผลลัพธ์ตามไกอาร์เรล กรอบการปฏิบัติตามกฎระเบียบ และร่องรอยการตรวจสอบ ซึ่งเป็นสิ่งที่เรื่องราวอย่างนี้ทำให้จำเป็น

ดูวิธีการทำงาน →