A firewall can enforce the boundary around a network while also becoming an attacker's route into it. Check Point's 12 November security advisory describes that sequence: attackers exploited PHP2Shell through the Mobile Access portal, then used a separate Gaia vulnerability to gain root privileges. The vendor reports exploitation from 30 October through 4 November 2026, after a PHP2Shell hotfix became available on 23 October.
The immediate concern is not simply the severity of one bug. It is the combination of an exposed application, a local privilege boundary and the network access entrusted to a security appliance. A corrected software version closes a vulnerability; it does not establish what happened while the device was exposed.
Two flaws crossed two different boundaries
PHP2Shell refers here to CVE-2026-109618 and CVE-2026-109619. The vendor's advisory describes unauthenticated remote code execution through affected gateway web interfaces. In the observed chain, that provided initial execution through Mobile Access. CVE-2026-123039 then supplied the escalation to root.
The second flaw is local. Check Point's R82.10 Take 52 release notes describe a root privilege escalation in the DynamicCLI daemon through a local world-writable control socket. A socket is a communication endpoint; making this one available to less-privileged processes undermined the intended separation between those processes and a privileged service.
That mechanism is not evidence of a second unauthenticated internet-facing listener. An attacker first needs the local access or execution required to reach it. In the documented incidents, the portal compromise supplied that prerequisite. Calling the DynamicCLI flaw local therefore describes its entry condition, not the maximum business impact of a chain that includes it.
The Canadian Centre for Cyber Security's 12 November advisory identifies exposure across multiple versions of Check Point Force Firewalls, Management and Log Servers, Multi-Domain systems and Maestro Orchestrator. It also records the vendor's confirmation of exploitation. Inventory review consequently needs to include management infrastructure, rather than only the public VPN address.
What the ESXi finding does and does not show
Check Point Research's original investigation describes webshells and a persistent TSGA backdoor with an embedded TSProxy networking component. In the investigated intrusion, the attacker moved from the gateway to a VMware ESXi host over SSH and deployed a wiper.
The researchers distinguish an initial 30 October wave, which included destructive activity against some Israeli organizations, from a broader 2 November wave without evidence of destruction. They do not establish a high-confidence attribution or motive.
This is evidence of a path from a compromised gateway into internal infrastructure. It is not evidence that CVE-2026-123039 directly exploits ESXi, escapes a virtual machine or grants control over every server behind a firewall. The observed SSH pivot depended on access beyond the appliance itself. Those additional relationships are part of the incident's significance: a perimeter failure can expose credentials and reachable services whose authority extends much further than the original portal account.
For a hosting operator, separating guest workloads from hypervisor administration is therefore relevant even when the first exploited product is a network appliance. The published case demonstrates the consequence of a successful pivot, not a universal outcome for all Check Point deployments.
Fixed releases have branch-specific identifiers
The November release notes document fixes for both PHP2Shell CVEs and CVE-2026-123039 in these Jumbo Hotfix releases, all dated 12 November:
A Take number belongs to its release branch. Take 52 on R82.10 and Take 14 on R82.20 are not comparable as a single numeric sequence. Nor should this list be read as a complete compatibility statement for every Check Point product or older release. The applicable product bulletin remains the authority for an individual deployment.
CERT-FR's 13 November notice independently flags privilege escalation and security-policy bypass across affected Check Point products. It refers to three vendor bulletins. Their grouping does not mean all three vulnerabilities were used in the observed chain: Check Point's advisory identifies exploitation for CVE-2026-123039 while listing no known in-the-wild exploitation for CVE-2026-107484 and CVE-2026-120490 at publication.
Remediation status and incident status need separate evidence
An update answers whether the running software contains a correction. It cannot, by itself, answer whether an intruder previously created a webshell, obtained credentials or reached another management service. Those are different questions with different evidence requirements.
Our OpenSSH and GPUStack analysis examines the same distinction after confirmed exploitation. The Check Point case adds a concrete appliance-to-hypervisor example. It also differs from the NetScaler November update, where the cited vendor disclosure did not establish exploitation of the newly announced issues.
Check Point cautions that an absence of matching indicators does not rule out compromise. For an exposed gateway, a defensible closure decision therefore needs both the applicable corrected release and an assessment of the earlier exposure window, persistence and downstream access. A successful maintenance ticket establishes only the first of those.

