PHP2Shell Flaws Put Internet Facing Systems at Risk, Patch Releases Arrive

| 2 Min Read
Two PHP flaws can enable unauthenticated code execution. Teams need to map exposure, install vendor fixes and check for signs of access.

PHP’s reach turns two flaws in its request handling into an infrastructure problem: vulnerable code may sit in public websites, internal tools and embedded products alike. The PHP Foundation released fixes Wednesday morning after disclosure of the bugs, which can allow unauthenticated remote code execution. For organizations, the immediate challenge is finding every affected deployment before attackers do.

One PHP flaw can hide in many kinds of products

PHP runs on the server side of websites and services including WordPress, Mailchimp, Drupal, Magento and Laravel, as well as parts of Slack and Zoom. WordPress powers 43% of the web, while a Forrester analyst cited PHP as underpinning 70% of websites. Those figures point to the scale of potential exposure, but a website inventory alone won't capture the full picture.

The Netherlands National Cyber Security Centre’s description of the flaws, collectively called PHP2Shell, highlights PHP’s presence in appliances, network devices and management interfaces. Some products bundle PHP behind their own interfaces, leaving administrators dependent on the product vendor to provide a fix. CERT-EU likewise warned that commercial and open-source products may be affected even when they don't directly expose a PHP installation.

The two flaws have CVSS severity scores of 9.2. CVE-2026-109618 can lead to remote code execution through memory corruption during out-of-memory handling, a process triggered when RAM and swap space are exhausted. CVE-2026-109619 involves stale, request-owned global state being reused by a later request on the same worker.

Both issues affect memory management and error handling, areas that nearly every request touches, the PHP Foundation said. That broad reach helps explain why the defects are difficult to address without compatibility or performance tradeoffs. Researchers at OX Security independently reproduced the vulnerabilities and found that the affected functionality is enabled by default and has been present since PHP 5.4.

Find exposure beyond the corporate website

“Checking the corporate website alone is not enough,” said Madelein van der Hout, a senior analyst at Forrester. Organizations need to locate PHP across internet-facing systems and products, then prioritize the systems most accessible to attackers. The PHP Foundation says all installations serving web requests may be at risk, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed. It identifies mod_php as the highest-risk configuration because an unauthenticated request may trigger exploitation before application code runs.

The Foundation calls immediate patching “imperative.” Affected installations should update to PHP 8.5.12, 8.4.27, 8.3.36 or 8.2.35, available through the PHP downloads page. Linux and long-term support vendors received fixes in advance, so teams should check their vendor’s guidance and packages.

PHP 8.1 and older end-of-life releases no longer receive fixes from the PHP project. Some distributors and LTS vendors still provide backported patches, however, making vendor status worth checking even for systems that can't move to a current PHP branch. Organizations using the volunteer-maintained Remi RPM repository were advised by the Foundation to switch to distribution packages or build from source to get the updates.

If an update can't be applied immediately, CERT-EU describes preventive actions to reduce exposure. Those measures don't make a vulnerable installation safe, so teams should treat them as temporary risk reduction while they work toward a vendor fix. The Foundation’s guidance and vendor advisories should inform decisions for systems where PHP is bundled or managed outside the organization’s direct control.

Availability decisions are part of the response

Some government websites in the Netherlands and Finland, along with services run by other organizations, went offline as a precaution after the disclosures. Rumors of publicly available exploits have also circulated. A shutdown doesn't establish that a system was compromised; it can be a deliberate choice when a team cannot patch or adequately protect an exposed service.

Van der Hout described temporary downtime as a proportionate risk decision: an organization weighs a short interruption against the possibility of a serious compromise. That decision depends on more than the patching team. It tests whether the business can tolerate a service outage and whether the right people have authority to take a system offline.

Patch installation also doesn't answer whether an attacker already gained access. Teams should investigate for signs of compromise after applying fixes, as van der Hout emphasized. “A patch fixes the vulnerability; it does not necessarily remove access an attacker may already have established,” she said. The response therefore needs to cover both remediation and incident investigation.

AI is shaping both discovery and response

AI tools featured in the vulnerability discovery and response. OpenAI said the red-teaming organization Calif identified a critical flaw using GPT-5.6 Sol through its Patch the Planet initiative. Wiz independently discovered and reported both vulnerabilities, while OpenAI partners provided research, technical support and coordination before PHP’s disclosure. OpenAI has also published a PHP version checker as a Codex skill to help teams identify vulnerable releases and select updates.

OpenAI called the issue one of the most impactful security problems in recent memory, given PHP’s severity and reach. The company framed early AI-assisted discovery as a “defender’s window”: maintainers had time to prepare fixes before attackers could exploit the flaws at scale. But the same capabilities can help attackers develop tools and exploits. OpenAI’s account of its role is available in its disclosure post.

Matt Saunders, DevOps lead at The Adaptavist Group, warned that automated systems may scan for vulnerable targets as soon as a flaw becomes public. He also said attackers could combine these critical vulnerabilities with other, lower-severity weaknesses to gain access. OX Security CEO Neatsun Ziv put the exposure in blunt terms: “When a flaw hits technology this ubiquitous, the blast radius isn’t one product or one industry. It’s the web itself.”

The practical takeaway is to treat PHP as a dependency that may be buried across the estate, not just as a web application runtime. Map where it runs, prioritize internet-facing deployments, apply the matching fixes and investigate for signs of access. Where those steps can't happen quickly enough, temporary downtime may be a sound operational decision.

Source: William Martinez · www.csoonline.com

Comments

Please sign in to comment.