Blog / Front Line — part 20 of 25

Check Point gateways backdoored via PHP2Shell: what to check

· urgentic · 5 min read

Short read

This post covers attacks on unpatched Check Point Security Gateways, in which attackers used the PHP2Shell flaws to get in through the Mobile Access portal, took root control of the gateway, installed a hidden backdoor and, in some cases, wiped a VMware ESXi server.

You'll learn what happened, why a compromised firewall is so hard to spot, and what to check today if you run Check Point gateways, even if you've already patched.

Check Point has confirmed that attackers exploited PHP2Shell on its Security Gateways to plant a persistent backdoor, and in some cases went on to destroy data. The gateway is meant to be the thing that protects your network, so when it's compromised, the attacker sits in the one place most organisations trust completely. If you run Check Point gateways with Mobile Access turned on, this deserves your attention today.

What happened

In a security advisory published on 12 November, Check Point said it had seen PHP2Shell exploited in the wild against its gateways, together with a local privilege escalation flaw in Gaia OS, tracked as CVE-2026-123039. It says the activity ran from 30 October to 4 November, that fixes had been available since 23 October, and that "only a couple of dozen unpatched systems, in a single region, were compromised" in the incidents it observed. Affected customers were contacted directly.

Check Point Research's technical write-up describes two waves. The first, on 30 October, targeted Israeli organisations and included destructive activity against some of them. A broader second wave on 2 November reached organisations in several countries and sectors, with no evidence of destructive activity.

We wrote about PHP2Shell itself when it was made public in October, in PHP2Shell: patched isn't safe until you check for compromise. This is what it looks like when it's used against a security appliance.

How the attack worked

According to Check Point Research, the attack followed a clear chain:

  1. Initial access through the Mobile Access portal. The attackers exploited PHP2Shell (CVE-2026-109618 and CVE-2026-109619), which allows unauthenticated remote code execution through the gateway's web interface.
  2. Root on the gateway. They then exploited CVE-2026-123039, a missing authorisation check in the Gaia OS DynamicCLI service, to go from the portal's service account to root.
  3. Web shells and a backdoor. With root, they placed PHP web shells in the directory the portal uses to build its login pages, then launched a backdoor called TSGA, which renames itself slot_CPconfigdaemon to look like a genuine Check Point component.
  4. Hidden access through normal services. TSGA carries a second component, TSProxy, which is injected into selected appliance processes. It picks out specially marked attacker connections while letting ordinary traffic carry on as normal.
  5. Moving inwards. The attackers ran a scanner on the gateway to map internal systems, then logged in to a VMware ESXi host over SSH using the vpxuser service account.
  6. Destruction. On the ESXi host, they ran a wiper called destroy_v6, which overwrites and deletes files, and left notes named DREAM_ANNEXATION.txt.

The Hacker News reports that TSGA's capabilities include configuration theft, persistence through backups and snapshots, and tampering with the firewall's own logs.

Who is behind it

Check Point says TSGA and TSProxy are modified versions of PhiliKit, a malware framework that ESET has linked to a China-nexus group. But it's careful not to name the attacker. The wiper and the political message in the notes don't fit that group's usual behaviour, and Check Point lists several possibilities: the same actor ran the whole campaign, it only provided initial access to someone else, it was a false flag, or the destruction was a smoke screen. For defenders, the attribution matters less than the method, which will be copied.

Why a compromised gateway is so hard to spot

Most monitoring treats the firewall as a source of truth. Its logs tell you what crossed the network edge, and it's rarely the thing being investigated. This attack turned that on its head:

  • The backdoor looks like part of the product. A process named after a Check Point daemon won't stand out to someone glancing at a process list.
  • Attacker traffic hides inside normal traffic. TSProxy handles the attacker's connections inside existing appliance services, so there may be no obvious new listening port.
  • The logs themselves can't be fully trusted. Check Point lists "unexplained gaps in firewall traffic logs, particularly following a crash of the logging daemon" as an indicator.
  • The next step uses a legitimate account. An SSH login to ESXi with a real service account looks routine unless someone knows it shouldn't come from the gateway.

That's why patching on its own isn't enough. A gateway that was compromised before it was patched can keep its backdoor afterwards.

What to check today

Based on Check Point's advisory and research, if you run Check Point Security Gateways:

  1. Install the November Jumbo Hotfix on all Management Servers and Security Gateways, starting with internet-facing gateways that have Mobile Access enabled.
  2. Hunt for the indicators Check Point has published, including:
    • PHP session files containing the string upload_progress_pwn
    • a process or file using the name pattern CPconfigdaemon
    • scanning or wiper tools under /tmp, such as files with "scanner" or "wiper" in the name
    • a note referencing "Dream Annexation" with no ransom demand
    • connections from the scanning addresses 130.94.19[.]204 and 130.94.19[.]168
  3. Look for gaps in your firewall logs since late October, especially after a logging daemon crash. Compare them with logs from other sources, such as your SIEM, endpoints or upstream network devices.
  4. Check SSH access from the gateway to internal systems, particularly ESXi hosts, and review use of the vpxuser account.
  5. Treat a match as a full incident. Check Point is clear that "absence of a match does not rule out compromise, and presence of a match warrants immediate investigation". If you find signs of compromise, assume configuration and credentials held on the gateway have been taken, and rotate them.
  6. Make sure ESXi backups are offline or isolated, so a wiper on the hypervisor can't reach them.

If an MSP or MSSP manages your firewalls, ask them directly whether your gateways had Mobile Access exposed, when the hotfixes were applied, and whether they've checked for the indicators above.

The wider lesson

This is the latest in a run of attacks on the devices that sit at the network edge: firewalls, VPNs and remote access portals. They're attractive because they're exposed to the internet, highly trusted and rarely watched as closely as laptops and servers. The practical answer is to treat them as systems that can be compromised: patch them fast, collect their logs alongside evidence from endpoints, identity and the rest of the network, and have someone looking for the moment the gateway starts behaving like a foothold rather than a guard.

You can read more about how urgentic approaches round-the-clock detection and investigation on our platform page.

  • Check Point
  • PHP2Shell
  • CVE-2026-123039
  • firewall
  • edge devices
  • backdoor
  • wiper
  • ESXi