Check Point Gateways Wiped: PHP2Shell Grew Up Into Espionage and a VMware ESXi Wiper
Check Point Gateways Wiped: PHP2Shell Grew Up Into Espionage and a VMware ESXi Wiper
PHP2Shell has moved from a bug class to a campaign. Check Point's own research team now reports that threat actors exploited the two critical PHP flaws — CVE-2026-109618 and CVE-2026-109619, fixed in PHP's emergency patches on October 21 — against unpatched Security Gateway appliances, first against Israeli organizations starting October 30, then in a broader wave on November 2 hitting organizations across multiple countries and sectors. CISA had already added both flaws to its KEV catalog on October 26, and Check Point says it observed active exploitation against "a couple of dozen customers," with evidence of successful compromises in some cases.
The kill chain is a fully assembled playbook. Initial access came through the gateways' Mobile Access portal using PHP2Shell, followed by privilege escalation via a newly disclosed zero-day in the DynamicCLI interface, CVE-2026-123039 — a flaw CISA added to KEV on November 13, days after the attacks were already observed. With root, the actors deployed PHP web shells, then the TSGA backdoor (masquerading as slot_CPconfigdaemon), which embeds a networking component called TSProxy that injects into legitimate appliance processes to listen for attacker connections concealed inside ordinary traffic. Post-exploitation included the open-source gogo scanner for internal discovery, then a pivot to VMware ESXi hosts over SSH using the vpxuser service account — and on the ESXi layer, the destructive finale: a Linux ELF wiper called destroy_v6 that zero-fills and recursively deletes data, accompanied by a politically themed DREAM_ANNEXATION.txt note that explicitly rejects ransom negotiation.
Attribution is deliberately cautious. Check Point tentatively links TSGA and TSProxy to PhiliKit, a toolkit ESET documented in June 2026 and attributed to the China-nexus Armory Castle cluster (formerly UNC5221) — but the wiper focus, the Israel-first targeting, and the explicit Israel–Palestine political framing don't match the espionage profile, so the company declined to link the campaign to any known actor. What matters operationally: the November Jumbo Hotfix fixes both PHP2Shell exposure and the DynamicCLI escalation on gateways and management servers, and detection is tractable — session files named upload_progress_pwn, unexpected CPconfigdaemon processes, gaps in firewall logs, and unauthorized SSH from gateways toward internal ESXi hosts are all concrete, searchable signals.
What Should You Do?
- Install the November Jumbo Hotfix on all Management Servers and Security Gateways, prioritizing any gateway with an internet-facing Mobile Access portal — that portal is the entry door this campaign walked through.
- Hunt retroactively on every previously exposed gateway: upload_progress_pwn session files, CPconfigdaemon binaries and processes outside expected paths, firewall-log gaps, and SSH sessions from the gateway to ESXi hosts that nobody initiated.
- Rotate the ESXi vpxuser credential path if the gateway could reach your hypervisor layer — the campaign authenticated over SSH with that service account, so a compromised gateway means treating the ESXi tier as exposed.
- Assume wiper, not just spyware. This campaign destroyed data with no ransom demand, so recovery readiness (tested restores, ESXi-level immutable backups) is the control that decides the outcome, not negotiation playbooks.
The WAF Angle
The uncomfortable lesson is that the web tier of a firewall product is an attack surface like any other — a pre-auth RCE in the PHP engine behind the Mobile Access portal gave root on an appliance whose entire job is stopping attacks. That's the the-guardian-has-a-web-app pattern, and WAF operators should internalize it: management portals, remote-access UIs, and "internal" admin endpoints on security appliances deserve the same WAF coverage, virtual-patch rules, and exposure discipline as customer-facing apps. A rule blocking the PHP2Shell request pattern on an exposed portal would have broken this chain at step one, and KEV timing shows CISA agreed — both CVEs were catalogued within five days of the fix. The same portal-as-perimeter risk sat behind our PHP2Shell coverage when the flaw class first emerged and Barracuda's pre-auth root chain in its email gateway — security appliances are web applications, and pretending otherwise is how gateways become beachheads.