PHP2Shell: Critical PHP Vulnerabilities Trigger Emergency Patches and Exploitation Warnings

Última actualización: 10/29/2026
  • PHP2Shell refers to CVE-2026-109618 and CVE-2026-109619, two critical PHP remote code execution flaws fixed in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35.
  • CISA added both bugs to its Known Exploited Vulnerabilities catalog and told civilian agencies to patch by Thursday, while advising inventory, offline treatment or temporary WAF coverage where patching is not immediate.
  • Attack attempts, scanning and fingerprinting were widely reported, but early public evidence did not confirm a successful compromise; some vendor claims were retracted and researchers kept validating working exploits.
  • A separate PHP flaw, CVE-2026-115560, has no patch yet and affects only unusual configurations; the suggested mitigation is to disable session upload progress.

PHP vulnerabilities security update

PHP keeps a huge part of the web running, so a critical security notice for the language tends to travel fast. In this case, the PHP security team scheduled an emergency release for Wednesday, 21 October 2026 at 14:00 UTC, saying it would fix two remote code execution vulnerabilities rated 9.2 under CVSS v4.0. Administrators were told to assume that any PHP installation they manage could be exposed until patched version numbers and technical details were published.

The two bugs became widely known as PHP2Shell, tracked as CVE-2026-109618 and CVE-2026-109619. The project shipped fixes in PHP 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and the message to users was blunt: update as soon as possible. The flaws matter because many PHP deployments sit behind public websites, APIs and web applications, and exploitation can lead to code running on the server without a normal login step.

fallos críticos en PHP
Related article:
Critical PHP Flaws Trigger Emergency Patches and Outages

What PHP2Shell actually affects

The risk is concentrated in PHP-based web servers and worker processes, including common setups built around Apache, php-fpm, PHP-CGI and similar application stacks. Attackers do not need a valid user account if they can reach a vulnerable service and trigger the right request path, which is why defenders were told to treat internet-facing PHP services as high-priority assets. The issue is not limited to one CMS or one framework; it follows the language runtime and the way applications are deployed.

Patching sounds simple, but the operational reality is messier. Some teams run PHP inside containers, some rely on distributions that backport security fixes, and others inherit old versions from hosting panels or third-party appliances. The fixed releases provide a clear target, yet administrators still need to confirm which PHP binary or package their applications actually use. A server can appear updated at the control panel level while a separate PHP worker or container image remains on an older build.

Palo Alto Networks confirmed that PAN-OS and Panorama customers were affected and said fixed PAN-OS releases had not yet shipped at the time of its statement. That detail matters because network security appliances are often assumed to be outside the reach of a language-level flaw. In practice, any product that embeds PHP or exposes a PHP-backed component can inherit the same risk, and vendors have to ship their own corrected builds even after upstream PHP patches are available.

software governance con inventario de tecnologías alojadas
Related article:
Software governance with hosted technology inventory: tools and strategy

Exploitation warnings and the evidence gap

CISA added the two vulnerabilities to its Known Exploited Vulnerabilities catalog on Monday evening and ordered civilian federal agencies to patch them by Thursday. The agency declined to describe the exploitation or identify victims, but its KEV rules say active exploitation can include both attempted and successful attacks. It also urged organizations to inventory PHP systems, apply the update, take PHP-based services offline if patching is not possible, and consider temporary web application firewall coverage as a stopgap.

Public reporting painted a tense but mixed picture. Several security companies said they saw scanning, fingerprinting and requests using non-working exploits, including attempts that began before the patches went live. Amazon Threat Intelligence said attackers started targeting the flaws four days before the release, and Cloudflare, watchTowr, Previdian, Hadrian and Patchstack were among those tracking unsuccessful activity. Dataminr’s threat research team said observed activity remained limited to scanning and non-working exploits, with no known compromises at the time of its report.

Some early claims of successful exploitation were posted and later retracted, which is a useful reminder that exploitation reports need evidence, not just a noisy payload. Experts had already tempered talk of mass compromise because a reliable proof-of-concept was not publicly available at first. That does not make the bugs less serious; it means defenders should separate confirmed incidents from attempts, scans and speculative claims. The phrase active exploitation can cover failed attempts, so a KEV listing does not automatically prove that attackers have won.

Related article:
Solved: html filter file upload

The exposure numbers are still hard to ignore. PHP is used by roughly 70% of websites, and VirusTotal estimates cited in reporting put more than 5.1 million PHP hosts online, with about 3.8 million having at least one port open to attack. Censys created a PHP2Shell dashboard showing 42 million internet-facing PHP services, of which more than 10 million had been patched since 21 October. Dozens of governments and businesses took sites offline last week, and researchers warned that millions of old PHP installs may never be updated because the applications around them are too fragile or too custom.

Scanners, dashboards and detection clues

Searchlight Cyber released php2shell-scanner, an open-source utility that checks whether a host is exposed to CVE-2026-109618 and CVE-2026-109619. Such tools are useful for quick triage, but they are not a substitute for asset inventory or patch verification. A scanner can tell a team that a service answers in a suspicious way; it cannot prove that every virtual host, background worker or container on the same machine has been properly remediated. Results should be tied back to the actual application owner and the package source.

Detection teams also have process-level clues to work with. A hunting approach for PHP post-exploitation can look for unexpected child processes spawned under Apache, php-fpm, PHP-CGI or other web workers, especially shell launches, downloads with curl or wget, chmod commands in temporary directories, reverse-shell constructs, encoded commands, persistence attempts and access to cloud credentials. The query behind that logic is not specific to these two CVEs and can catch other web application exploits, so it will produce false positives in busy hosting environments. Even so, unusual process trees under a web worker deserve a look when a critical PHP bug is under attack.

The same monitoring can help with containment. If a web worker suddenly spawns a shell, pulls a file from an external host, makes a script executable in /tmp or reads cloud metadata, the response should move from patch management to incident response. Teams should preserve logs, check for web shells and review outbound connections rather than simply restarting the service. This is especially important because public scanning activity can hide targeted attempts among thousands of failed requests.

Related article:
Solved: referer url

A separate flaw with no patch yet

While the PHP2Shell response was still unfolding, the PHP team disclosed another remote code execution issue, CVE-2026-115560, with no patch available at the time. The bug only applies to systems with an unusual combination of configuration and application code: an application must set session.serialize_handler to php_binary or php_serialize, then switch it to php at runtime with ini_set() before calling session_start(), and session upload progress must still be enabled. Applications that never change the serialize handler at runtime can keep the feature on, according to the advisory.

Because that setup is unlikely in practice, the project said it would handle the flaw in its regular release cycle and publish a full advisory with the next release. Until then, the recommended mitigation is to set session.upload_progress.enabled=0. The PHP team noted that administrators who had already applied that setting during the PHP2Shell preparation should leave it in place. The key point is that this is not a blanket emergency for every PHP site; it is a narrow configuration edge case, but one that still deserves checking in custom applications.

Discovery, response and operational reality

The discovery story adds another layer. OpenAI said the PHP flaws fit its Patch the Planet initiative for open-source security, and Califf found one of them using the GPT-5.6 Sol model. Wiz independently found and reported both bugs. OpenAI also proposed a five-stage process for handling critical vulnerabilities in widely used software, based on its two-month response to the PHP issues. The details are less important for most administrators than the outcome: AI-assisted bug hunting is now part of the disclosure pipeline, and vendors need faster ways to validate and communicate fixes.

For organizations that cannot patch immediately, the advice has been consistent: know what you run, reduce exposure, and watch for exploitation. CISA pointed to PHP guidance and advisories from Australia, Canada, New Zealand, the United Kingdom and other governments. Web application firewalls may buy time, but they are not a permanent fix, and they can miss requests that do not match known patterns. Taking a vulnerable service offline is painful, yet it is often safer than leaving a publicly reachable PHP runtime exposed while a working exploit matures.

The practical to-do list is still familiar. Inventory every PHP runtime and container image, verify the patched release, remove or isolate end-of-life deployments, enable the session upload progress mitigation where the newer flaw could apply, and monitor web workers for suspicious child processes. The PHP2Shell alarm may have started with two CVEs and a rushed release, but the longer lesson is about basic exposure management: old PHP, forgotten virtual hosts and unclear ownership are exactly what turn a language-level bug into an operational problem.

php
Related article:
Critical PHP php2shell flaws force urgent patching and WAF rollouts
Related posts: