PHP2Shell’s biggest enterprise risk may be the systems that security teams can’t patch on demand. The flaws affect PHP, a runtime embedded in web servers, appliances, and operational technology devices. Even organizations that identify vulnerable systems may have to wait for vendors to prepare fixes—or manage products that will never receive one.
Two critical vulnerabilities were announced last week in PHP. Tracked as CVE-2026-109618 and CVE-2026-109619, the flaws are known as PHP2Shell. Attacks are underway: on Monday, CISA added both vulnerabilities to its Known Exploited Vulnerabilities catalog. Researchers had already observed widespread scanning and attempted exploitation, though they had not confirmed production compromises at that point.
The exposure reaches beyond conventional web servers. Network-attached storage devices, firewalls, VPN gateways, printers, building management systems, industrial IoT equipment, and human-machine interfaces can all run web interfaces powered by PHP. In many cases, the runtime is bundled into a product’s software or firmware, leaving customers dependent on the manufacturer’s update schedule. Some devices also run PHP versions that have reached end of life and no longer receive upstream security fixes.
That combination gives PHP2Shell a different kind of long tail from Log4Shell. “With Log4j, the main problem was finding a library you didn’t know you had. With PHP, much of the forgotten infrastructure will be found, but won’t have an upstream fix,” Sai Molige, senior manager of threat hunting at Forescout, told CSO via email.
Log4Shell offers a warning about how long exposure can persist. A year after the flaw in the widely used Log4j Java library was disclosed in December 2021, nearly three quarters of organizations still had vulnerable assets. Log4j also remained a prominent attack vector in real-world incidents throughout 2023. PHP2Shell could present an even harder cleanup for organizations whose PHP runtimes are inside products they can’t update themselves.
Finding the runtimes—and the fixes
The PHP project lists PHP 8.5.12, 8.4.27, 8.3.36, and 8.2.35 as fixed releases. But those upstream version numbers don’t tell the whole story. Linux distributions package and maintain PHP independently, and long-term support releases often ship older branches. Their maintainers may backport a security fix without adopting the latest upstream version number, making simple version checks unreliable.
The support picture varies by distribution. PHP 7.4, 8.0, and 8.1 have reached end of life, but the PHP developers published minimal source backports to help package maintainers address the flaws in those branches. PHP 7.3 and older received no such coverage. VulnCheck found that 42% of the 2.6 million internet-facing PHP hosts it scanned ran end-of-life branches: 27% used PHP 3.x through 7.3, while another 15% ran PHP 7.4, 8.0, or 8.1.
Whether a distribution has a fix can depend on its release and support arrangement. Ubuntu 22.04 offers a PHP 8.1 update, while the PHP 7.4 fix for Ubuntu 20.04 requires a Pro subscription. Debian 11’s PHP 7.4 fix is available through a Freexian extended-support contract. AlmaLinux 8 provides backports for retired PHP 7.2, 7.3, and 8.0. Red Hat Enterprise Linux, SUSE, Alpine, and Amazon Linux each define their own support boundaries.
Those backports can confuse scanners that rely on reported version strings. AlmaLinux’s patched package, php-8.3.33-1.el10_2.alma.1, has a lower version number than PHP’s upstream fixed release for that branch, 8.3.36. A scanner that compares only the banner against the upstream number could therefore mark a fixed host as vulnerable.
Visibility is harder still when PHP lives in prebuilt containers, virtual machines, firmware, or third-party software bundles. “The problem for most companies is they don’t know where they’re running PHP,” said Eylam Milner, co-founder and CTO at Echo. “It’s in base images and in components that bundle their own runtime, and a lot of it is on versions older than 8.2 that are past end of life.”
There is no public count of enterprise appliances and IoT or OT devices running vulnerable PHP, although some are internet-reachable. Flare analyzed service records from multiple internet scans, covering 2,220,231 records from PHP hosts across 224 countries. In a regional sample from one US state, the firm found NAS devices, VPN gateways, remote access portals, building automation systems, and physical access controllers.
Cogent Security’s scan of dozens of Fortune 500 customer environments on the disclosure day found an average of 2,266 affected assets per company. Of those, 926 were actively running the PHP service, and 122 were internet-exposed. The assessment covered servers and virtual machines, not hardware appliances or OT equipment. The gap points to an inventory problem across asset types, rather than a shortage of scanning tools.
“The real battle isn’t just the vulnerability; it’s visibility,” said Tom Kellermann, vice president of AI security and threat research at TrendAI. Teams will need to combine package-manager data, network scans, and software bills of materials where available. Each method sees a different slice of the environment, and none guarantees coverage of runtimes hidden inside appliances or vendor images.
Reducing exposure while vendors prepare updates
A complete PHP inventory still leaves teams with product-specific decisions. Some vendors may take weeks to deliver firmware or software updates; industrial devices can remain in service for a decade or longer. In the meantime, manufacturers may recommend disabling a management interface, restricting access, or isolating a device from other network segments.
Vendor responses already vary. Palo Alto Networks scheduled updates for some PAN-OS devices for this week. Siemens told customers using its RUGGEDCOM APE1808 industrial networking devices, which are also powered by PAN-OS, to follow Palo Alto’s advisory. Delta Controls patched its enteliWEB enterprise building management platform, while firmware fixes for affected HVAC and room controllers were still pending. Some NetModule IIoT and industrial routers from Belden were confirmed vulnerable, with patches in development.
For systems that can’t be updated promptly, immediate steps include moving the vulnerable web application off exposed network segments, placing firewall controls in front of it, and checking for signs of prior compromise. If administrators don’t need HTTP or HTTPS management, disabling the web server may reduce exposure while preserving access through another protocol. Belden and Delta have advised customers to disable embedded HTTP servers; NetScaler recommends limiting its management GUI and API to trusted administrative networks.
Web application firewalls can add a layer of detection, but PHP2Shell’s multipart upload behavior creates blind spots. A WAF may inspect only part of a request or fail to parse its form structure deeply enough to catch every variation. An attacker could pad a request with junk data to push malicious content past an inspection limit. A direct connection to a backend server could also bypass a WAF entirely.
Cloud providers and security vendors have published rules for the flaws. AWS, Cloudflare, and Microsoft Azure offer WAF protections, and enterprise WAF products have added detections. Network intrusion prevention systems can help too, provided they can inspect encrypted traffic. Available signatures include Palo Alto Threat ID 543004, Check Point protection CPAI-2026-15172, Cisco Talos Snort rules 67359 and 67360, and Suricata rules 3000650 through 3000652.
Check for evidence of earlier access
Mitigation closes paths attackers might use next; it does not show whether someone already got in. Steve Povolny, vice president of AI strategy and security research at Exabeam, said PHP2Shell moved from scanning activity to confirmed remote code execution on a honeypot within days of the October 21 patches. “Patching closes the door. It doesn’t tell you who is already inside.”
That makes incident review part of the response, especially for systems that remained exposed after the patches appeared. Security teams can examine web server access and error logs, PHP-FPM records, WAF and reverse-proxy logs, CDN data, endpoint and process telemetry, DNS and outbound connection logs, authentication records, file-integrity alerts, and cloud audit trails.
Potential request indicators include PHP_SESSION_UPLOAD_PROGRESS, a PHPSESSID cookie, multipart/form-data sent to endpoints that don’t normally accept uploads, unusually large or malformed multipart bodies, deeply nested field names, repeated POST requests from one source, and a spike in HTTP 500 errors.
Host telemetry may reveal PHP-FPM, Apache, or httpd worker processes launching shells or utilities such as sh, bash, curl, wget, and base64, or starting unexpected interpreters. Investigators should also look for reverse shells, outbound connections from web workers, and new or changed PHP files in web roots, upload locations, and plugin directories. Xint.io researchers also point to abnormal PHPSESSID cookie lengths or corruption, unusually large session files, and unexplained worker restarts or segmentation faults.
PHP2Shell’s practical lesson is that patch status depends on more than the PHP version shown by a server. Enterprise teams need to know which products contain the runtime, which vendor owns the fix, and whether the system shows signs of earlier exploitation. For embedded systems that can’t be updated quickly, that inventory and exposure review will shape the response long after the first patches arrive.