A firewall is supposed to be the thing attackers can't get past. In this Check Point campaign, it became the thing they got through first, and then used as a base for everything else.
Check Point has released emergency hotfixes for several products after attackers exploited a high-severity privilege escalation flaw, CVE-2026-123039. The bug gives an attacker root privileges and the ability to send commands to the Gaia OS Dynamic CLI daemon, a core background process on the appliance. The vendor's research write-up details the exploitation. Hackers have been using the 7.8-rated flaw since late October, and Check Point says "a couple of dozen" customers were affected.
The initial foothold came from a different problem entirely. Attackers got in through PHP2Shell, the collective name for two critical PHP flaws tracked as CVE-2026-109618 and CVE-2026-109619. Those bugs are rated 9.2 and allow unauthenticated remote code execution. An attacker who can reach a vulnerable web portal doesn't need credentials. They just need the portal.
Frank Dickson of Dickson Research put it bluntly: "For customers, one foothold delivered the full menu: Spying, staying, and destroying. For Check Point, the cost is trust. A firewall is supposed to be the strongest point in the network, not the softest."
How a local bug becomes a full takeover
CVE-2026-123039 sits in the Dynamic CLI inter-process communication interface, which passes privileged command requests to the root-level Gaia OS Dynamic CLI. It's a local privilege escalation flaw, so an attacker has to already run code on the appliance. That's why it scores 7.8 and not higher. But the PHP2Shell chain removes that barrier, so the local requirement stops mattering much.
The sequence ran in two steps. Attackers first reached the Mobile Access portal through the PHP2Shell bugs. They then escalated to root using the CVE-2026-123039 exploits, which were zero-days at the time.
The root step is the part that should worry anyone running this class of appliance. Dickson described the core problem: a management service runs as root, listens on a local socket that any process on the box can write to, and never checks who is calling. Any code that lands on the device can talk to a service that already has the keys. That design gap is what turns a web bug into a full-appliance compromise.
Once they had root, the attackers went deep. They hid web shells inside the templates the portal uses to build its login page. They also installed a TSGA backdoor and TSProxy malware. According to Dickson, that malware survives reboots, copies itself into backups and snapshots, and hides among the appliance's own processes. Standard cleanup habits, like restoring a backup, can carry the infection forward.
The backdoor can also do two things that matter for anyone trying to map the network. It can copy the firewall configuration, which Dickson called "essentially a blueprint of the network." It can alter the firewall's logs, which hides what the attacker did.
From the gateway to the hypervisor
Check Point reported that after taking root, the attackers pivoted from the compromised gateway to a VMware ESXi host over SSH. They deployed the destroy_v6 wiper and left politically themed messages signed by a group calling itself Dream Security Ltd. That's the point where the incident stopped being an intrusion and became a destructive event.
The trust relationship between a firewall and a hypervisor is the detail that made this possible. Management paths are often allowed through because they're convenient and assumed to be safe. A gateway that can reach the virtualization layer is only as secure as its most exposed component, and here that component was a portal sitting on the internet.
Check Point's guidance is to deploy the CVE-2026-123039 hotfixes right away, starting with devices that expose Mobile Access to the internet. Teams that can't patch immediately can stop the dynamic_server process and keep it from restarting, though the vendor warns this disables some CLI commands. Affected products include the Firewall, Multi-Domain Security Management Server, Security Gateway, Security Management Server, and Scalable Platforms.
Dickson's additional advice is more pointed. Turn off Mobile Access wherever it isn't used. Keep management interfaces off the internet entirely. Run the advisory's log checks on every gateway exposed since late October. Then cut the path between the firewall and the hypervisor, because a potentially exposed gateway should never have a trusted route into ESXi. Ship firewall logs off the box as they're generated, to a system the firewall can't reach or modify.
Security tools are now the high-value target
Erik Avakian, a technical counselor at Infotech, framed the problem as one of trust inversion. Once an attacker controls a gateway, the device becomes a launchpad against systems that would otherwise stay protected. The tool built to defend the network is now the most useful thing to attack.
"Organizations can no longer assume their security infrastructure is inherently secure simply because it's designed to provide protection," Avakian said. Firewalls, VPN appliances, and security gateways draw attackers because they offer the most leverage and control. He argues these devices need the same access control, vulnerability management, monitoring, and incident response discipline as other critical infrastructure.
His practical checklist starts with aggressive patching of internet-facing systems and restricted management access. It adds monitoring for unusual administrative activity and hunting for signs of persistence or lateral movement. He also flagged a trap many teams fall into. Patches don't remove attackers who are already inside. Affected systems need investigation and, where necessary, a rebuild from trusted images.
Avakian also pointed to AI as a pressure multiplier. With AI speeding up attacks, he said CISOs should build a zero trust approach across identity, network, applications, data, and devices together, rather than treating each pillar as a separate project.
Edge devices sit where monitoring is thinnest
Scott Miserendino, CTO at DataBee, made the structural point. Edge devices are primary targets because they sit at the boundary, hold privileged access, and rarely get the endpoint monitoring other systems receive. Security appliances, in his view, should be handled as the high-risk assets they've become.
His advice starts from an assumption many teams resist: assume you've already been breached. Hunt for published indicators of compromise, review network and endpoint logs routinely, and segment networks to limit blast radius. "Security tools aren't automatically secure, so defend the defenders," he said.
Dickson added that PHP2Shell isn't going away soon. PHP sits underneath a large share of web software, including the portals bolted onto security appliances. Defenders should treat it as a standing risk, not a one-time patch cycle.
The management plane is the prize
Check Point isn't the only vendor where attackers have targeted firewall management. The pattern is industry-wide.
In January, Fortinet disclosed that attackers abused its FortiCloud single sign-on to log into other customers' FortiGate and FortiManager devices and create administrator accounts. In July and September, CISA added Check Point management server flaws to its catalog of exploited vulnerabilities. That same month, it also added a Cisco Secure Firewall Management Center flaw, rated 10.0, that gives an unauthenticated attacker root access. In October, Palo Alto Networks warned of attacks against the PAN-OS management interface, with root on the firewall as the prize.
"The logic is simple," Dickson said. "The management plane holds the rulebook for the entire network."
Switching vendors looks like an escape hatch, but Dickson argues it mostly swaps one weak point for another. Every edge device has a web portal, and every web portal is a door. "The remedy is fewer doors," he said. "Shrink the internet-facing edge and put remote access behind one consistent identity policy."
That's the takeaway I'd carry into any firewall review this quarter. A firewall is not a wall. It's a computer on the internet with privileged access to everything behind it. Attackers are already treating it that way, and defenders need to start doing the same, by reducing what's exposed, separating management from production, and shipping logs somewhere the device can't touch.