The PHP2Shell vulnerabilities, tracked as CVE-2026-109618 and CVE-2026-109619, expose servers that process web requests in PHP to a critical remote code execution risk. An unauthenticated request can reach the PHP engine before the site or application code even runs. PHP lists several affected environments, including PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed.
Patches are available in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35. Organisations must still verify the versions actually in use, including in products and appliances that embed PHP.
The timeline sharpens the urgency: AWS observed an exploitation attempt as early as 17 October 2026, before the public disclosure on 21 October. That does not prove every attempt succeeded, but it reminds us that an IT security flaw can be exploited before every team has identified its vulnerable systems.
What is PHP2Shell and why do these flaws matter for businesses?
PHP2Shell is the name given to two flaws discovered in the PHP core — the language that powers dynamic pages and web requests to a server. They therefore do not target only one specific application, plugin or CMS. The risk sits in the PHP engine when it receives and processes web requests.
That widens the perimeter to check. A WordPress site may be affected because it runs on PHP, but WordPress is not necessarily the cause of the flaw. Other websites, business applications, admin interfaces and web services also rely on this language. PHP can also be embedded in third-party products, appliances and network equipment. In those cases, the organisation does not always see which version is running behind the scenes.
The flaws can allow a remote, unauthenticated attacker to execute code on the vulnerable server. Application access control is therefore not always enough to block the risk: in some configurations, PHP processes the request before application code can intervene.
To know which systems and services to review, start with the cybersecurity practice and a cybersecurity audit. The first step is to locate PHP environments, including those managed by a host or embedded in a vendor product.
Both vulnerabilities share this general risk, but their technical mechanisms differ.
CVE-2026-109618: memory corruption triggered by an OOM error
The first vulnerability concerns how PHP handles certain memory errors while processing a request.
CVE-2026-109618 is linked to an OOM (out of memory) error, when PHP can no longer allocate the memory needed for an operation. To interrupt processing in that situation, PHP notably uses a mechanism called longjmp(). The problem appears when that interruption occurs while some internal request data is not yet fully initialised.
PHP must then clean up resources used by the interrupted request. In some circumstances, that cleanup phase may treat memory information as valid when it is not. The corruption is therefore not limited to a crash or an error message: it can open the door to remote code execution on the server, with the privileges of the PHP process.
The key point for IT teams is that the scenario sits inside PHP’s internal processing. It does not necessarily depend on a flaw in the site’s business code. An application that appears to work normally can therefore remain exposed if its PHP environment has not received the patch. Configurations that serve web requests directly should be reviewed, especially when they are reachable from the Internet.
This flaw relates to poorly managed memory state during an interrupted request. The second, CVE-2026-109619, instead concerns data that may persist from one request to another.
CVE-2026-109619: memory state reused across requests
CVE-2026-109619 concerns internal request-related data that may remain present after the corresponding memory has been freed. A fatal error can leave this stale state in a PHP process. If that process then handles another request, it may reuse that information before its internal state is correctly reset.
This behaviour matters especially in environments where processes or workers stay alive and handle several requests in sequence. The issue does not mean every site visitor automatically receives the previous visitor’s data. It means that, under certain conditions, PHP may rely on memory state that should no longer be used. That mismanagement can enable remote code execution.
CVE-2026-109619 is therefore distinct from CVE-2026-109618. One is about reusing a stale internal state; the other is about memory corruption during the handling of an out-of-memory error. Both affect PHP, can be triggered without authentication, and do not necessarily require a separate application flaw.
That difference also explains why a temporary measure aimed at one particular processing path is not enough to fix the whole problem. To assess exposure, you now need to look at which PHP versions and execution modes are actually in use.
Which servers, versions and products may be affected?
PHP has released patched versions for its maintained branches: 8.5.12, 8.4.27, 8.3.36 and 8.2.35. Earlier versions of those branches should be treated as exposed until the vendor has confirmed an appropriate fix. For PHP 8.1 and older branches, the PHP project no longer provides regular updates. Some distributions or vendors do, however, ship backported patches. Check their exact status before concluding that a legacy system is protected.
The risk does not depend on a single server type. Affected environments include PHP-FPM, Apache with mod_php, php-cgi, FrankenPHP and LiteSpeed. A server may host several PHP versions in parallel: an update can therefore fix one site without fixing other applications or pools.
The check must also cover products that ship their own PHP runtime. That may include appliances, admin interfaces, business software or container images. Updating the CMS does not guarantee that the server’s PHP engine is patched. Likewise, the version shown by a command-line call may differ from the one used by the web server.
If hosting is managed by a provider, ask for the active version for each web service and the reference of the installed fix. Version numbers alone are not always decisive: a vendor may have backported a patch without changing the main PHP version number.
Once the inventory is ready, the next question is whether vulnerable systems have already been targeted.
Is PHP2Shell already being exploited?
AWS reports spotting an exploitation attempt on 17 October 2026 — four days before the two CVEs and the patches were published. After the 21 October disclosure, proof-of-concept material circulated quickly and observed attempts multiplied. AWS notes, however, that observing an attempt does not prove it succeeded on a real server.
Researchers at Palo Alto Networks also reported attacks against Internet-exposed admin interfaces running affected PAN-OS versions. Their telemetry referred to more than 22.6 million potentially exposed instances. That figure describes systems identified as likely vulnerable; it does not mean 22.6 million compromised servers.
These elements justify treating PHP2Shell as urgent for exposed systems. They do not, however, allow you to claim that every vulnerable install has been attacked, or that a detected attempt necessarily led to an intrusion. Analysis should rely on logs and the systems concerned, and take into account that suspicious activity may have preceded the official publication.
Organisations that find a vulnerable version should therefore quickly check whether it has been updated, then look for signs of abnormal activity. The priority remains patching the PHP engine, because temporary settings do not fully close these flaws.
After identifying the environments involved, the goal is to confirm that the fix has actually been applied on every server, site and third-party product.
How to check exposure and apply the patches
Start by listing the services that process PHP requests. Include websites, business applications, admin interfaces, staging servers, containers and appliances. A CMS-only inventory is not enough: several sites may use different PHP versions or processes on the same hosting.
For each environment, record the version used by the web server, not only the one shown in a terminal. The published fixed versions are PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35. If an application runs on an older branch, ask the vendor whether a backported fix exists and how it is identified. For a product that embeds PHP, consult the vendor’s security advisory.
On managed hosting, ask the host to confirm in writing the active version and the update date. On infrastructure you administer yourself, update the packages or images in use, then restart or replace the PHP processes concerned. A patch installed on disk does not guarantee that older in-memory processes have been refreshed.
After deployment, re-check the version actually being served, application behaviour and any forgotten instances. A cybersecurity audit can fold this check into a broader review of assets, configurations and vulnerabilities — see also the cybersecurity hub.
If an immediate update is impossible, some measures can reduce exposure. They do not, however, fix the vulnerabilities.
What temporary measures can you take if the update is delayed?
Temporary settings vary by flaw and configuration. For CVE-2026-109618, recommendations notably cover memory management and inbound requests. Some administrative settings can prevent the application from restoring an exploitable memory limit. You should also constrain request size at the web server or proxy layer, taking forms and file uploads into account. These changes can affect site behaviour and should be assessed before deployment.
For CVE-2026-109619, one measure is to disable session.upload_progress.enabled in a configuration loaded before request processing, then restart the PHP processes concerned. On PHP 8.5, disabling fatal_error_backtraces can also reduce some known paths. These settings do not necessarily block every possible variant of the vulnerability.
A web application firewall or reverse-proxy filtering can add temporary protection. It does not replace the PHP patch. Also avoid improvised system-limit changes: some memory-limiting mechanisms can trigger precisely the errors involved in CVE-2026-109618.
Each measure should be validated for the environment concerned, monitored after deployment and paired with an update plan. The cyber attack protection page can complement this with a broader risk-reduction approach. If compromise is possible, you must also move from prevention to investigation.
Even after installing a patch, an organisation may still need to check whether a server was compromised before it was updated.
What if compromise is suspected?
Preserve web server, PHP, system and security-tool logs before any cleanup. They can help reconstruct events. Because attempts were observed before public disclosure, do not limit analysis to the CVE publication day alone. Review periods preceding the update in particular.
Look for unusual requests, processes launched by the web server, recently created or modified files, and unexpected connections. Requests containing the string PHP_SESSION_UPLOAD_PROGRESS may be worth examining, but their presence alone does not prove exploitation. Conversely, the absence of that string is not enough to rule out compromise.
If serious indicators appear, limit exposure of the system concerned and contact the host or a competent incident-response team. Avoid immediately deleting suspicious files or reinstalling the server before preserving material useful for analysis. After the update, review accounts, secrets and access that may have been exposed, then restore the service from a known-good state.
Preparation also matters: an IT security policy should state who to alert, which logs to keep and how to decide to isolate a service. A penetration test can then assess a defined perimeter, but it does not replace analysis of an already suspected incident.
PHP2Shell is therefore a reminder that a technical vulnerability must be handled within the wider framework of asset and incident management.
PHP2Shell: key takeaways
PHP2Shell does not only concern teams that develop in PHP. Any organisation running a site, application or appliance on a vulnerable engine version must check its exposure. Both CVEs can enable unauthenticated remote code execution, and some attempts were observed even before the patches were published.
The priority is therefore to identify the PHP environments actually in use, then confirm that each has received a fix. The versions published by PHP are 8.5.12, 8.4.27, 8.3.36 and 8.2.35. Older branches require verification with the vendor, especially when a patch has been backported or PHP is embedded in a third-party product.
Temporary settings and network protections can reduce some risks, but they do not replace the update. Organisations should also review logs and files on their systems if earlier exposure is possible. That approach assumes you know which assets exist, who administers them and how to react when an alert appears.
PHP2Shell thus shows the importance of continuous vulnerability management: inventory, check vendor advisories, apply patches, verify deployment and retain what is needed for investigation. That follow-up turns an IT security flaw into a concrete operational risk — not just a technical alert.
Frequently asked questions
Short answers on PHP2Shell, fixed versions, WordPress, hosting and temporary measures.
Does PHP2Shell affect WordPress sites?
A WordPress site can be exposed because it runs on PHP. The vulnerability affects the PHP engine, not WordPress alone. Check the version used by your hosting, as well as other applications or products that may embed their own PHP environment.
Which PHP versions fix CVE-2026-109618 and CVE-2026-109619?
The published fixed versions are PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35. For older branches, check with your host, vendor or supplier whether a backported patch has been installed.
Is a user account or password enough to block these flaws?
Not necessarily. PHP processes some requests before application code runs. Site authentication mechanisms therefore do not replace updating a vulnerable PHP engine.
What if my host manages PHP?
Ask which version is actually used for each site and service, whether the patch is applied and whether PHP processes have been restarted. A terminal command may show a different version from the one that handles web requests.
Is a temporary configuration enough to fix PHP2Shell?
No. Some settings or filters can reduce exposure paths, but they do not fully fix both vulnerabilities. An update to a patched version, or a vendor-confirmed fix, remains necessary. See also our cyber attack protection page.