PHP2Shell Exploit Attempts Hit Millions as Patch Race Escalates

Exploit attempts against the PHP2Shell flaws surged into the millions, with Cloudflare recording attacks before public disclosure.

@sharedot

PHP2Shell Exploit Attempts Hit Millions as Patch Race Escalates

By @sharedot · · 8 pages

  • Programming
  • Php
  • Php2shell
  • Security

Exploit attempts against the PHP2Shell flaws surged into the millions, with Cloudflare recording attacks before public disclosure.

What happened: from thousands to millions of attacks

The PHP Foundation shipped emergency patches for CVE-2026-109618 and CVE-2026-109619 — collectively nicknamed PHP2Shell — on October 21, and exploit attempts exploded almost immediately.

According to Cloudflare CEO Matthew Prince, the company saw thousands of attempts to exploit the flaw in the hours before it was publicly announced, which jumped to tens of thousands within 45 minutes of the announcement, hundreds of thousands within two hours, and millions now.

Kaspersky reports the same trajectory, noting attempts reached tens of thousands within the first 45 minutes after disclosure and climbed to millions by October 22.

The alarming part: attacks started before disclosure

The most unsettling detail is that the exploitation race began before defenders even knew the bug was public.

Startup Fortune reports that exploitation attempts were logged as early as four days before the vulnerabilities were formally disclosed, and that details leaked online before the PHP Foundation's intended release window — meaning attackers and defenders learned about the bug at effectively the same moment.

Oman Observer, republishing The New York Times' reporting, adds that Amazon revealed in a blog post that an unknown actor had attempted the attack against its honeypot network four days before the patch's release, with Amazon CISO C.J.

Moses saying it remains unclear how the actor acquired exploit information prior to disclosure.

An AI model found a bug that sat undisturbed for 14 years

The bugs had lurked in PHP's code for 14 years, and it took AI to prove them exploitable.

According to Startup Fortune, researcher Ryan Austin at the security firm Calif used OpenAI's GPT-5.6 Sol model in August to comb PHP's source code under the 'Patch the Planet' initiative, finding a partial version of the flaw; OpenAI's own team found a fuller version in September, and Wiz independently found both without knowing of either effort.

Kaspersky reports PHP developers had first noticed signs of one vulnerability in a 2024 audit but could not demonstrate remote code execution and postponed a fix.

Wiz called the findings 'among the most severe vulnerabilities discovered in PHP, and arguably in software history,' saying proving exploitability required a chain of steps 'at the edge of what human researchers could practically achieve.'

What the flaws actually let attackers do

Both vulnerabilities score 9.2 out of 10 on the CVSS v4.0 severity scale, which CERT-EU says signals unauthenticated remote code execution — an attacker can run commands on a vulnerable server over the internet with no password or account.

Startup Fortune explains the technical roots: CVE-2026-109618 is a use-after-free triggered when PHP runs low on memory mid-request, while CVE-2026-109619 comes from stale global state, where a PHP worker holds references to freed memory that a later request reads.

Kaspersky reports patched versions landed as 8.5.12, 8.4.27, 8.3.36 and 8.2.35, and Cloudflare says its web application firewall was blocking exploit attempts across every customer plan, including the free tier, starting at 3:02 a.m. UTC on October 21 — roughly nine hours ahead of public disclosure.

The stakes: 70% of the identified web

PHP's footprint is what turns two memory bugs into a global event.

Multiple outlets report that PHP runs roughly 70% of all websites whose server-side language can be identified, per W3Techs, and it powers WordPress, which itself runs about 40% of all websites, from blogs and shops to government portals and news sites.

Government, school and hospital websites across Europe pulled systems offline as a precaution, Spectrum News reports, and Oman Observer notes Dataminr had seen no confirmed successful attacks as of Thursday — only scanning, fingerprinting and non-working exploits — though it called successful intrusions 'only a matter of time.'

OpenAI called it 'one of the most impactful security issues in recent memory.'

What comes next: a shrinking patch window

The working takeaway for developers is that disclosure-to-exploitation time is collapsing.

Kaspersky notes approximately 2,700 critical vulnerabilities were discovered worldwide in Q2 2026, a 34% increase year over year, partly attributed to AI making source-code analysis far more efficient — and the same tools can help attackers develop exploits faster.

Its guidance is to patch PHP to a secure version immediately, restrict public network access where patching is impossible, deploy a web application firewall, and audit software versions including rarely touched services.

Startup Fortune frames the harder problem for CTOs: the uncomfortable number is 70%, because PHP likely hides in your vendors, CMS plugins and SaaS dependency chain, and this week showed how little warning you get to find it.

Sources

  1. omanobserver.om › What is PHP? Why are government and business websites going offline?
  2. kaspersky.com › Websites taken offline over critical PHP vulnerabilities | PHP2Shell
  3. startupfortune.com › An AI model found a PHP bug so bad it took government sites offline
  4. en.prothomalo.com › What is PHP and why did it take government websites offline?
  5. spectrumlocalnews.com › Global PHP security flaw prompts safety measures for N.Y. government websites

More on Programming